Two dates in the EIA weekly retail gasoline series can break a bid: a 1990-91 collection gap and a May 14, 2018 methodology change. A fuel price is really a value attached to a product, a place, and a date range, and any of those three can be wrong even when the number looks reasonable. The first starts a stretch with no weekly gasoline data. The second is a change in how the weekly gasoline estimates were produced. Either one can make a series look continuous when it is not.
Key takeaways
- Pick the exact series your bid needs: fuel, region, time unit, and period.
- Check whether your bid or baseline dates fall inside the December 10, 1990 to January 14, 1991 gap.
- Check whether any comparison crosses May 14, 2018, when the methodology changed.
- Do not use an EIA monthly or annual average for a period with incomplete weekly data.
- Record the page address, the pull date, and the first and last week in your extract.
This post walks through the checks in the order you would normally make them: pick the right series, pull it and keep a record, check the two dates, check the averages, and write down what you used. Each step is short. The point is to make sure the number that goes into the bid is the number you think it is.
1. Decide exactly which series your bid needs
Start with the bid, not the data. A common error with fuel inputs is pulling a number that is close to the right one but describes something else: a different product, a different region, or a different time unit. Before you open anything, write down four things. The first is the fuel, gasoline or diesel. The second is the geography, national or a regional breakdown. The third is the time unit, a single week, a month, or a year. The fourth is the period the bid actually depends on, whether that is a current price, a historical baseline, or an average over a job's duration. For a wider look at fuel inputs, see equipment fuel cost and gasoline price figures.
It helps to know what the EIA's weekly report is. In its explainer on gasoline prices, the agency describes the report this way: "Gasoline and Diesel Fuel Update—This weekly report provides the most recent available and historical nominal prices for retail gasoline nationally and by PADD region." ( EIA, Gasoline prices and outlook)
That one sentence settles three things that matter for a bid. The report is weekly, so it is a series of weekly values and not a daily quote. It covers retail prices, which are what buyers pay at the pump, so it is not a wholesale or contract price. And it reports prices nationally and by PADD region, which means geography is a real choice you have to make on purpose. A national figure is the wrong input for a job in one region, even if the national figure looks plausible.
The same sentence says the prices are "nominal." In plain terms, that means they are not adjusted for inflation. For a current bid this rarely matters. For a comparison across several years, it does. If your bid uses a historical figure to justify a current one, state whether the comparison is in nominal or adjusted terms, and do not mix the two in one table.
Note also the product question. The weekly retail page is titled "U.S. Gasoline and Diesel Retail Prices," so both fuels appear on it. But the two notes covered in this post, the collection gap and the methodology change, are written about weekly retail gasoline. If your bid uses diesel, do not assume those notes apply to it. Read the diesel rows and the page notes on their own terms before you rely on either date.
2. Pull the weekly series and keep the dates attached
Go to the EIA page titled "U.S. Gasoline and Diesel Retail Prices" at eia.gov/dnav/pet/pet_pri_gnd_dcus_nus_w.htm. Pull the weekly series you chose in step one, national or regional and gasoline or diesel, and save it. For more on using these prices in a fuel line, see the guide to EIA gasoline prices for fuel lines.
Save it in a form that keeps each value next to its week date. A price with no date attached cannot be checked against a bid window, cannot be lined up against the EIA's own notes, and cannot be shown to a reviewer later. Plain text copied from a web page often loses the date column or shifts it by a row, so check that the first and last values match the page before you go further. If you paste into a spreadsheet, keep the date column as real dates, not as text, so sorting and filtering behave as you expect.
Record three facts alongside the values. The first is the page address, so anyone can see where the figures came from. The second is the date you pulled the data, so anyone can see what the series showed on that day. The third is the first and last week in your extract, so anyone can see whether the extract covers the period the bid depends on. It is easy to skip the pull date, and then nobody can rebuild which figures a bid used when a client or auditor asks about them months later.
Once the file is saved, do not overwrite it. If you pull the series again for a revised bid, save the new pull as a new file with its own date in the name. The earlier file stays as it was, and the bid can point to the version it used. This is the simplest way to keep the inputs of a bid fixed while the work around it changes.
Finally, look at the shape of the data before you use any of it. Scan the week dates from start to finish. Look for blanks, repeated dates, and gaps. A blank or a jump in the dates is exactly what the next steps are about, so do not hide it by sorting, filling, or interpolating at this stage. Note it and move on. You will decide how to handle it once you know what it is.
3. Check the 1990 to 1991 collection gap
The first date to check is a gap in collection rather than a change in method. The EIA's weekly retail page states: "We did not collect weekly retail motor gasoline data between December 10, 1990, and January 14, 1991." ( EIA, U.S. Gasoline and Diesel Retail Prices)
That sentence is about gasoline, and it names one window with a start and an end. For a bid, the consequence is direct. If the period you need falls inside that window, the weekly gasoline series has no value for those weeks. Check how your downloaded file shows those weeks before you rely on it.
The most common way this gap breaks a series is a silent join. Suppose you copy a column of weekly values into a sheet, sort it, and calculate a change from one row to the next. The row before the gap and the row after it end up next to each other, and the calculation treats the two weeks as if they were consecutive. A chart drawn from the same data will look smooth across the gap, which makes the problem harder to see. The same thing happens when a formula averages the rows in a range that quietly includes the gap, or when a lookup finds the nearest available date and returns a value from a different week without saying so.
Check the gap in three places. First, the bid's own reference period, if the bid relies on historical prices. Second, any baseline you use to measure change, such as a year-earlier comparison, a prior-period average, or a rolling window. Third, any formula that counts weeks. A formula that divides a total by a fixed number of weeks will give the wrong answer across the gap, because the total has fewer values behind it than the divisor assumes.
If the gap does touch your period, you have several options, and choosing among them is a judgment for the estimator and, where the contract requires it, the client. You can exclude the missing weeks and say so in the bid. You can use only the weeks on one side of the gap if that side covers what the bid needs. You can mark the affected figure as unavailable rather than produce a number that looks complete. What you should not do is fill the missing weeks silently with neighboring values and present the result as the EIA's series. If you do fill them for a reason, label each filled value and say what it was filled from.
Also check the edges of the gap against your bid dates, not only the middle. A bid window that starts or ends a few days inside the gap can still be affected, because the first or last available week may be far from the date the bid actually uses. Compare the exact dates, and do not round them to the nearest month.
4. Check the May 14, 2018 methodology change
The second date is a change in how the figures were produced. The EIA's weekly retail page states: "On May 14, 2018, we implemented statistical methodology changes to improve the accuracy of the weekly retail gasoline price estimates." ( EIA, U.S. Gasoline and Diesel Retail Prices)
Unlike the gap, this change leaves values in the series. Nothing is missing. The problem is that the values on one side of May 14, 2018 were estimated with one method and the values on the other side with a different one. The change was made to improve accuracy. Do not estimate how large its effect was for any week and present that estimate as the EIA's correction. A series that crosses that date was produced under two methods.
That matters most when you measure a movement rather than read a level. If you calculate a percentage change, a trend, or a seasonal comparison across May 14, 2018, part of what you see may come from the change in method rather than from the market. A reader who only sees the final percentage will not know that. A reviewer who knows the history of the series will ask.
To handle it, first check whether your bid's window or your baseline crosses May 14, 2018. If it does not, you have no issue from this change. If it does, you have two reasonable paths. You can split the series at that date and measure each side on its own, so that no comparison crosses the change. Or you can keep the full span and state in the bid that the comparison crosses a methodology change. The first approach is cleaner when the thing you are measuring can be defined on each side. The second is more honest when the bid needs one continuous comparison and the split would make the result meaningless.
Do not assume the note applies to diesel because it appears on the same page. The quote is about the weekly retail gasoline estimates. If your bid uses diesel, read the diesel series and its notes directly, and record what you found, including if you found nothing relevant.
The table puts both dates and the averaging rules from the quotes in one place. Its first two columns carry the facts as the EIA states them, and its last column is the practical reading of each rule for a bid.
5. Check monthly and annual averages before you use them
Some estimators skip the weekly values and use the monthly or annual figure, because it is simpler to put in a bid. That shortcut is only safe if you understand how the figure was built. The EIA's weekly page says: "Monthly and annual averages are simple averages of the weekly data." ( EIA, U.S. Gasoline and Diesel Retail Prices)
That sentence has two consequences. The first is that a monthly average is a function of which weeks it contains. If a month has a full set of weekly values, its average is built from all of them. If you build your own average for a month with fewer weekly values, it no longer represents the same thing as a full month. The second is that a simple average weights each week equally. It is not a volume-weighted or a time-weighted figure, and it does not account for how many days a week covers within a month. If your bid depends on a monthly figure as a weighted cost, a simple average is a different quantity, and you should say which one you are using.
The same page also says: "For months and years with incomplete weekly data series, the monthly and/or annual averages are not available." ( EIA, U.S. Gasoline and Diesel Retail Prices) So when the weekly series is incomplete for a month or a year, the EIA's average for that period should not be expected.
This applies directly to the gap from step three. The window runs from December 10, 1990 to January 14, 1991. That overlaps part of December 1990 and part of January 1991, so both of those months have incomplete weekly data in the sense the quote describes. Under that rule, you should not expect an EIA monthly average for either month, and the annual average for 1990 or 1991 should not be expected either. If your bid uses one of those periods, the honest answer is that no published average exists for it, and the bid needs a different basis.
If you build your own average from the weekly values, state that you did, and use the same method the EIA states: an equal-weight simple average of the weeks you have. Make sure the weeks you include are the ones you intend. A week that begins in one month and ends in another needs a rule, and the rule should be written down. Pick one, such as the month in which the week ends or the month in which most of its days fall, and apply it the same way every time. Do not mix a self-built monthly average with an EIA figure in the same table without labeling each one. A reader who sees two numbers under the same heading will assume they came from the same method.
Also keep the difference between an average and an estimate clear in your wording. The EIA figure is an average of its weekly data. A bid line that uses it is an estimate of future cost built on that average, and the bid should say so in its own terms without claiming the EIA made a forecast. The source describes what the average is. Your use of it is a separate decision.
Illustrative example: a 4-week window with one missing week
Suppose a bid window covers 4 weeks, and your extract has values for only 3 of them because one week falls in a period with no collected data. A self-built average for that window is the simple average of the 3 values, divided by 3, not by 4. Label it as self-built, list the 3 weeks it covers, and do not present it as a published EIA figure. If the window also crosses May 14, 2018, either split the comparison at that date or state the crossing in the bid.
6. Write a short record of the dates you used
The last step is a record that lets someone else rebuild the number. Keep it in the bid file next to the spreadsheet, not in an email thread or in your memory. A short record is usually enough, and it should answer the questions a reviewer is most likely to ask.
Include the series name and fuel, and the region if it is not national. Include the page address and the date you pulled the data. Include the first and last week in your extract. State whether the bid's window contains any part of the December 1990 to January 1991 gap, and if it does, say what you did about it: excluded the weeks, used one side of the gap, marked the figure unavailable, or filled values and labeled them. State whether the bid's window or baseline crosses May 14, 2018, and if it does, say whether you split the series or stated the crossing in the bid. Finally, say whether you used EIA monthly or annual figures or built your own averages, and if you built them, which method and which rule for weeks that span a month boundary.
That is six or seven lines. A record of this size is short enough that people will read it, and complete enough that the bid can be checked without you in the room. Keep it with the version of the bid that used it. For a related guide on dated price records, see diesel price date stamps. If the bid is revised and you pull the series again, add a new record for the new pull rather than editing the old one, so that each version of the bid still points to the inputs it actually used.
Finally, check the bid against any contract or client requirement that names a price source or an averaging method. If the contract specifies a series, a region, or a basis for escalation, that requirement controls, and the EIA notes in this post do not override it. The two dates covered here can break a weekly gasoline series without anyone noticing. A gap in the data and a change in the method do not look like errors on a chart. They look like ordinary weeks, which is why they are worth checking on purpose before the number goes into a bid.
Next step: today, open the EIA weekly page, pull the weekly series you chose in step one, and check its first and last weeks against your bid window.
Frequently asked questions
Does the 1990-91 gap affect every EIA gasoline series?
No. It covers only the weeks between December 10, 1990 and January 14, 1991, so check whether your bid or baseline dates fall inside that window.
Does the May 14, 2018 change make earlier prices wrong?
Only comparisons that cross it need care, because values before and after that date were produced under different methods.
Can I use an EIA monthly average for a month with missing weeks?
No. For months and years with incomplete weekly data, the EIA's monthly and annual averages are not available, so do not present a self-built average as an EIA figure.