Refinery Utilization Rate Explained

Refinery utilization is the gross feed to U.S. crude distillation units divided by their operable capacity. The EIA’s reading for the week ending October 16, 2026 was 93.4 percent, or 16,860 thousand barrels a day on 18,044 thousand of capacity. The 2025 annual average was 92.0 percent, and the range since 1990 runs from 78.9 to 95.6.

The percentage is a ratio of two numbers that move for different reasons. In the week ending October 9, operable capacity rose by 17 thousand barrels a day, which on its own lowers the rate by about 0.09 points even if no refinery changes its runs. A week later the Rocky Mountain district (PADD 4) printed 100.6 percent, which is possible because capacity is a rated figure, not a ceiling. Headlines that attach a story to a one-point move rarely check which half of the ratio moved, or which region moved it. The closures behind the denominator are covered in US refinery capacity in 2026, and the margin that decides whether plants want to run in refinery crack spreads.

The formula, one term at a time

The EIA glossary defines the refinery utilization rate as the gross input to the atmospheric crude oil distillation units divided by the operable refining capacity of the units. Both terms have definitions that matter.

• Gross input is the total feed to the atmospheric crude oil distillation units. The glossary lists crude oil, lease condensate, natural gas plant liquids, unfinished oils, liquefied refinery gases, slop oils and other liquid hydrocarbons. It is therefore larger than crude oil input alone.
• Operable capacity is the capacity that, at the start of the period, is in operation; not in operation and not under active repair but capable of being placed in operation within 30 days; or not in operation but under active repair that can be completed within 90 days. It is the sum of operating and idle capacity, measured in barrels per calendar day.
• Operating capacity, the part in operation at the start of the period, is a narrower measure and not the denominator in the headline rate.

Calendar-day capacity, in the EIA glossary, is what a facility can process under usual operating conditions, reduced from the maximum stream-day figure for limits such as downtime and the ability of downstream units to absorb output. The rate therefore compares what the plants ran with what they could have run on an average day. The weekly figure is published in the Weekly Petroleum Status Report on Wednesdays, on the schedule in the EIA release calendar guide.

Six weeks, recomputed by hand

The first habit to build is to check a printed rate against its parts. The table reproduces the EIA’s U.S. weekly inputs and utilization for the six weeks to October 16, 2026 and recomputes the ratio from gross inputs and operable capacity.

Every recomputed value rounds to the printed one. Two things to notice. First, operable capacity stepped up from 18,027 to 18,044 thousand barrels a day for the weeks ending October 9 and October 16. A seventeen-thousand-barrel change in the denominator is 0.09 percent, small, but it is the kind of revision that makes comparing weeks a few points apart less exact than it looks. Second, the numerator moves much more than the denominator: between the weeks ending September 11 and September 25, gross inputs fell from 17,449 to 16,670, a drop of 779 thousand barrels a day, while capacity stayed at 18,027. That is a 4.3 point fall in utilization from operations alone. The weekly series does not state a cause, but the regional pages, set out further down, show where the barrels went.

Crude input versus gross input

The same week offers an example of why the glossary definition matters. Crude oil inputs for the week ending October 16 were 16,649 thousand barrels a day, gross inputs 16,860: a gap of 211 thousand barrels a day of non-crude feed. If the rate used crude alone it would be 92.3 percent, not the 93.4 percent that the EIA prints. A figure quoted in a news story as “crude runs” or “crude throughput” is not the same quantity. When you compare utilization from two sources, check which numerator each uses.

Five regions, one national number

The national rate is the sum of regional gross inputs divided by the sum of regional capacities. The EIA publishes the five Petroleum Administration for Defense Districts (PADDs) separately; this site explains the regions in its PADD guide. For the week ending October 16, 2026:

The regional columns explain the national number. The Gulf Coast holds 54.9 percent of U.S. capacity, so its 95.9 percent drives the national rate more than any other region. The East Coast runs at 80.8 percent, far lower, but it holds only 5.1 percent of the capacity, so it pulls the national figure down by less than a point. A national reading in the low 90s can contain a Gulf Coast in the middle 90s and a coast at 80.

The Rocky Mountain row is the surprise. Its 657 thousand barrels a day of gross inputs exceed its 653 thousand of operable capacity, a utilization of 100.6 percent. Nothing in the definition prevents this: the denominator is a calendar-day rating of operable capacity, which already allows for usual limits on operation, and the numerator counts all feed to the distillation units, including non-crude liquids. A plant can run above its rated calendar-day figure for a week. A reading above 100 in a single region is therefore not an error in the data. It does signal that the plants were running harder than their rated average.

Where the September drop happened

The same six weeks are published for each district. Subtracting the week ending September 25 from the week ending September 11 shows which region carried the national fall of 779 thousand barrels a day.

The Midwest alone lost 619 thousand barrels a day, about four-fifths of the national fall, and its utilization went from 100.0 percent in the first week to 85.6 in the third. The Gulf Coast, which holds more than half of U.S. capacity, moved by only 104. A reader who looked only at the national line would see a broad weakening; the district pages show one region. They also show that the Rocky Mountain district went the other way, rising from 609 to 636 thousand barrels a day over the same two weeks. Which plants were involved is not something these pages say.

The ranges make the stability gradient visible: the Gulf Coast moved within a band of 1.9 points, the West Coast 3.2, while the Midwest swung 14.4 and the East Coast 11.3. The largest district moves the national average slowly; the smaller districts, where one plant can be a large share of the total, produce the weekly noise.

Thirty-six years of annual averages

The annual series sets a weekly number in context. It averages all weeks of a year, so it smooths out maintenance and storms.

Three features stand out. The 1990s climbed from 87.1 percent in 1990 to a peak of 95.6 in 1998, a decade average of 91.5. The years after the 2008 financial crisis were the weakest stretch before 2020, with 82.9 percent in 2009 and an average of 86.3 from 2008 through 2013. And 2020 stands alone at 78.9 percent, the lowest annual figure in the table, followed by a recovery to 86.6 in 2021 and 91.8 in 2022. From 2022 to 2025 the rate has been steady at 90.4 to 92.0 percent. EIA reports the 2025 gross input as 16,731 thousand barrels a day against 18,180 of operable capacity.

A different capacity base is part of why the history is not a straight line: the denominator has changed with closures and expansions. The annual rate is a ratio of two quantities that both moved over 36 years, so a comparison of 1998 with 2025 is a comparison of different industries, not a statement about efficiency.

The yearly calendar

Within a year, utilization follows a shape. The monthly series shows it for the last four full years and the months of 2026 so far.

The four-year mean is lowest in February (85.9) and January (87.2), climbs through spring to a summer plateau of 94.6 in June, 94.0 in July and 94.4 in August, falls in September (91.9) and bottoms again in October (88.7), before recovering in November and December. The October dip is visible in each year: 90.3 in 2022, 86.7 in 2023, 89.7 in 2024 and 88.1 in 2025. Scheduled maintenance outside the peak driving and heating seasons is the usual explanation for spring and fall troughs, though the series carries no causes of its own. Compare that with the weekly path above: the late-September weakness in 2026 sits in the same part of the calendar, and 2026 has already registered 96.5 percent in both June and July.

Six of the seven published months of 2026 ran above the four-year mean for that month, and April matched it. The gap was largest in January, at 4.2 points, and smallest in May, at 0.8. At current capacity the months above the mean are worth roughly 140 to 760 thousand barrels a day of additional gross input. A reader comparing this October’s 93.4 percent with the four-year October mean of 88.7 should expect the same direction: above average for the month, which describes how hard the plants have run in 2026 and not margins or prices.

A planning consequence is that a utilization number should be compared with the same time of year. A 93 percent reading in October is high for October and ordinary for August. The EIA’s own seasonal comparison in the weekly report works the same way.

Translating points into barrels

A percentage point means different volumes in different places. The table takes a hypothetical outage of 200 thousand barrels a day, an illustration and not a report of any event, and divides it by each region’s operable capacity from Table 2.

Nationally the same outage costs about one point. In the Gulf Coast it costs two. On the East Coast, where the whole district has 928 thousand barrels a day of capacity, it costs more than 21 points, which is why East Coast utilization is volatile and why a single plant’s status can dominate the regional figure. For the same reason, a national rate near 93 percent says little about whether any one region has room to add output.

The same conversion helps with the seasonal dip in Table 6. Using the EIA annual operable capacity for each year as an approximation, the drop from the August peak to the October low was about 683 thousand barrels a day in 2022 (3.8 points), 1,350 in 2023 (7.4), 789 in 2024 (4.3) and 1,327 in 2025 (7.3). Those are the volumes that moved out of the distillation units over two months in the fall, and the size varies by a factor of two between years. Because the series carries no causes, read the figures as a range for the seasonal swing, roughly 0.7 to 1.4 million barrels a day in these four years, not as a prediction for any one year.

What the rate does and does not tell you

• It measures use of distillation capacity. It says nothing about conversion units such as crackers and cokers, which determine how much gasoline and diesel a barrel yields, and nothing about whether products are being made at a profit.
• It does not fix prices. High utilization means there is less spare distillation capacity to add output. Whether prices respond depends on demand, inventories and crude cost, which the weekly report guide covers.

Data and sources: U.S. Energy Information Administration, U.S. Total Weekly Inputs & Utilization and the regional pages for PADD 3 and the other four PADDs, weeks ending September 11 to October 16, 2026, checked October 23, 2026 (crude oil inputs, gross inputs, operable capacity, percent operable utilization). EIA, U.S. Refinery Utilization and Capacity, annual, release of September 30, 2026, checked October 23, 2026 (2020 to 2025 gross input, operable capacity and utilization). EIA, Percent Utilization of Refinery Operable Capacity, annual and monthly series, checked October 23, 2026 (1990 to 2025 annual; monthly 2022 to July 2026). EIA, Glossary: refinery utilization rate, operable capacity, operating capacity, gross input, barrels per calendar day and barrels per stream day, checked October 23, 2026. Recomputed ratios, regional shares, the four-year monthly mean and the decade averages are this page’s arithmetic on those series.