U.S. railroads moved an estimated 113.5 million barrels of crude oil in 2025, about 311 thousand barrels a day, in EIA's movements-by-rail table (release of October 30, 2026). That is 30 percent of the 2014 peak of 382.0 million barrels, or 1,047 thousand barrels a day. The rail that stayed inside the United States, 81.2 million barrels, equals 1.6 percent of the 4,987 million barrels of U.S. field production, against 9.9 percent in 2014.
The fall was not even. Rail shipments that originated in the Midwest, PADD 2, where the Bakken sits, dropped from 261.1 million barrels in 2014 to 52.4 million in 2025, while the Rocky Mountain region, PADD 4, kept 59 percent of its 2014 volume. Rail from Canada peaked later, at 110.2 million barrels in 2019, and was 32.3 million in 2025. Through the first eight months of 2026 the monthly series is running 19.8 percent above the same months of 2025, led by U.S.-origin barrels.
What the EIA rail series counts
The rail series is an estimate, not a count taken at a loading rack. EIA describes the source of its table as "estimates of rail movements based on data from the Surface Transportation Board and other information." The Surface Transportation Board's Carload Waybill Sample is a sample of the waybills filed by railroads that end 4,500 or more revenue carloads a year, and EIA multiplies sampled carloads by the sampling rate, assigns each one to an origin and a destination PADD (or Canada), and converts cars to barrels. EIA stopped publishing its standalone "U.S. Movements of Crude Oil by Rail" page; the same series now sit on its Petroleum and Other Liquids data page and in the Petroleum Supply Monthly.
Table 1. How EIA builds the crude-by-rail series, from its table definitions and the March 2024 methodology document
| Item | What the documents say |
|---|---|
| Where the numbers come from | EIA calls them "estimates of rail movements based on data from the Surface Transportation Board and other information." The core input is the STB Carload Waybill Sample, submitted by railroads that terminate 4,500 or more revenue carloads a year. |
| What counts as crude | The EIA definition covers crude oil plus lease condensate, and the methodology states that bitumen and condensate are included in the crude-by-rail data. |
| Which movements | The table's own definition: "Crude oil movements on railroads to and from Canada, within the 50 States and the District of Columbia (including interstate and intrastate)." Moves inside one PADD are included. |
| Sampling | Waybill sampling rates ran from 1 in 2 carloads (trains of 101 or more cars) to 1 in 40 (single carloads) for 2010 to 2020; from 2021 they run from 1 in 2 to 1 in 5. |
| Carloads to barrels, 2010 to 2014 | Barrels per car from National Energy Modeling System production estimates by region, with a representative API gravity for each crude type, by quarter and origin PADD. |
| Carloads to barrels, 2015 on | Volume-weighted API gravity by PADD from the EIA-914 production survey, combined with the carload weight on the waybill. Canadian-origin cars use Canada Energy Regulator data. |
| Worked example in the methodology | Origin PADD 2, average carload 175,000 pounds, volume-weighted API 40: 607 barrels per car. |
| Months without sample data | The methodology says EIA "may estimate anywhere from one to three months of rail movements" with regression and averaging models, then replaces them as waybill data arrive. |
| History | EIA began publishing monthly crude-by-rail data in March 2015 and reports the series back to January 2010, so the 2010 to 2014 values were estimated retrospectively. |
Sources: EIA, Movements of Crude Oil by Rail, Definitions, Sources and Notes; EIA, U.S. energy-by-rail data methodology, March 2024 (file dated April 5, 2024; Tables 1, 15 and the sections on carload waybill sample data, converting to barrels and near-month estimates), all checked November 4, 2026. No newer edition was found at the EIA URL.
Three features of the method matter when you read the 2014 peak. First, EIA began publishing in March 2015, so the 2010 to 2014 volumes were estimated retrospectively, with barrels per car taken from modeled production rather than from the survey API gravities that EIA used from 2015; the peak year rests on the older conversion. Second, the sampling rate for waybills of one or two carloads was one in forty until the end of 2020 and one in five after it, which changes the precision of the small movements and leaves the 1 in 2 sampling of large unit trains alone. Third, EIA assigns each waybill to a month by adding four to sixteen days to the waybill date, depending on the type of train and the number of railroads, so a train that left on the last day of one month can be counted in the next.
The monthly and annual tables are one dataset: the twelve monthly values for each year from 2010 to 2025 add up to the annual figure to within 2 thousand barrels, which is rounding. EIA can revise rail data back to January of the second-most-recent year in any monthly cycle, and when the Petroleum Supply Annual comes out, usually in August, over a longer window. The figures here are the release of October 30, 2026; any later release can move them, and the newest months are the likeliest to change.
Annual volumes, 2010 to 2025
EIA reports the table in thousand barrels per year and splits it three ways: rail inside the United States (EIA's "Intra-U.S. Movements"), rail from the United States to Canada, and rail from Canada into the United States. Total rail was 23.8 million barrels in 2010 and 42.4 million in 2011, then 152.0 million in 2012 and 296.1 million in 2013, before reaching 382.0 million barrels in 2014. The three years 2013 to 2015 account for 37 percent of the 2,695 million barrels the table records over all sixteen years. The single biggest month in the monthly table was October 2014, at 35.3 million barrels, or about 1,137 thousand barrels a day.
Table 2. Crude oil by rail, 2010 to 2025, thousand barrels per year
| Year | Total | Within the U.S. | U.S. to Canada | Canada to U.S. | Total, thousand b/d | Total as % of 2014 |
|---|---|---|---|---|---|---|
| 2010 | 23,788 | 23,712 | 55 | 21 | 65 | 6.2% |
| 2011 | 42,370 | 41,371 | 256 | 743 | 116 | 11.1% |
| 2012 | 152,047 | 138,024 | 6,217 | 7,807 | 415 | 39.8% |
| 2013 | 296,064 | 259,194 | 7,391 | 29,479 | 811 | 77.5% |
| 2014 | 382,034 | 316,276 | 12,605 | 53,153 | 1,047 | 100.0% |
| 2015 | 317,446 | 275,218 | 4,414 | 37,814 | 870 | 83.1% |
| 2016 | 175,701 | 142,587 | 0 | 33,114 | 480 | 46.0% |
| 2017 | 139,389 | 87,053 | 102 | 52,234 | 382 | 36.5% |
| 2018 | 201,208 | 114,004 | 180 | 87,023 | 551 | 52.7% |
| 2019 | 251,224 | 141,015 | 0 | 110,209 | 688 | 65.8% |
| 2020 | 157,141 | 94,844 | 369 | 61,929 | 429 | 41.1% |
| 2021 | 126,064 | 65,070 | 0 | 60,994 | 345 | 33.0% |
| 2022 | 97,985 | 52,079 | 0 | 45,906 | 268 | 25.6% |
| 2023 | 97,329 | 65,191 | 0 | 32,139 | 267 | 25.5% |
| 2024 | 121,670 | 86,570 | 0 | 35,100 | 332 | 31.8% |
| 2025 | 113,505 | 81,207 | 0 | 32,298 | 311 | 29.7% |
Source: EIA, Movements of Crude Oil and Selected Products by Rail, annual, thousand barrels, release of October 30, 2026, checked November 4, 2026 (series ZAMN-ZAMN, NUS-NUS, NCA-NUS and NUS-NCA). The total equals the three parts in every year to within 1 thousand barrels. Thousand barrels a day and the percent of 2014 are calculated (annual total divided by 365 or 366 days).
Crude oil by rail, 2010 to 2025, million barrels a year
Rail that began in the United States (including exports to Canada) and rail that began in Canada.
- Began in the United States
- Began in Canada
Data behind this chart
| Year | Began in the U.S., thousand bbl | Began in Canada, thousand bbl | Total, thousand bbl |
|---|---|---|---|
| 2010 | 23,767 | 21 | 23,788 |
| 2011 | 41,627 | 743 | 42,370 |
| 2012 | 144,241 | 7,807 | 152,047 |
| 2013 | 266,585 | 29,479 | 296,064 |
| 2014 | 328,881 | 53,153 | 382,034 |
| 2015 | 279,632 | 37,814 | 317,446 |
| 2016 | 142,587 | 33,114 | 175,701 |
| 2017 | 87,155 | 52,234 | 139,389 |
| 2018 | 114,184 | 87,023 | 201,208 |
| 2019 | 141,015 | 110,209 | 251,224 |
| 2020 | 95,213 | 61,929 | 157,141 |
| 2021 | 65,070 | 60,994 | 126,064 |
| 2022 | 52,079 | 45,906 | 97,985 |
| 2023 | 65,191 | 32,139 | 97,329 |
| 2024 | 86,570 | 35,100 | 121,670 |
| 2025 | 81,207 | 32,298 | 113,505 |
Source: Table 2. EIA movements of crude oil by rail, annual, release of October 30, 2026, checked November 4, 2026.
The two sources of rail crude peaked in different years. Domestic rail and rail exports to Canada peaked in 2014; rail from Canada peaked in 2019 at 110.2 million barrels, more than double its 2014 level of 53.2 million. The Canada Energy Regulator ties its own 2019 peak in rail exports to export pipelines that were full and to Alberta's production limits. The lowest total after 2014 is 97.3 million barrels in 2023, the lowest of any year since 2011. Domestic rail fell from 316.3 million barrels in 2014 to 52.1 million in 2022, 84 percent lower, and has since recovered to 81.2 million, which is still 74 percent below the peak.
Rail against production and imports
Set rail against the supplies it draws from, and the 2014 peak looks smaller and the 2025 figure looks smaller still. U.S. field production rose 56 percent between 2014 and 2025, from 3,205 to 4,987 million barrels a year (the state detail is in US oil production by state), while domestic rail fell 74 percent. The ratio of domestic rail to production was 5.8 percent in 2012, 9.5 percent in 2013 and 9.9 percent in 2014, and 1.6 percent in 2025. The same exercise for Canada divides rail from Canada by U.S. crude imports from Canada: 5.1 percent in 2014, a high of 7.9 percent in 2019 and 2.3 percent in 2025. On a combined basis, total rail was 6.5 percent of production plus all crude imports in 2014 and 1.6 percent in 2025.
Table 3. Rail as a ratio of U.S. production and of imports from Canada, 2010 to 2025
| Year | U.S. field production, million bbl | Domestic rail, million bbl | Domestic rail as % of production | Imports from Canada, million bbl | Rail from Canada, million bbl | Rail as % of imports from Canada |
|---|---|---|---|---|---|---|
| 2010 | 2,002 | 23.7 | 1.2% | 719 | 0.0 | 0.0% |
| 2011 | 2,071 | 41.4 | 2.0% | 812 | 0.7 | 0.1% |
| 2012 | 2,388 | 138.0 | 5.8% | 888 | 7.8 | 0.9% |
| 2013 | 2,736 | 259.2 | 9.5% | 941 | 29.5 | 3.1% |
| 2014 | 3,205 | 316.3 | 9.9% | 1,052 | 53.2 | 5.1% |
| 2015 | 3,443 | 275.2 | 8.0% | 1,157 | 37.8 | 3.3% |
| 2016 | 3,242 | 142.6 | 4.4% | 1,181 | 33.1 | 2.8% |
| 2017 | 3,422 | 87.1 | 2.5% | 1,258 | 52.2 | 4.2% |
| 2018 | 4,005 | 114.0 | 2.8% | 1,353 | 87.0 | 6.4% |
| 2019 | 4,502 | 141.0 | 3.1% | 1,392 | 110.2 | 7.9% |
| 2020 | 4,154 | 94.8 | 2.3% | 1,316 | 61.9 | 4.7% |
| 2021 | 4,133 | 65.1 | 1.6% | 1,372 | 61.0 | 4.4% |
| 2022 | 4,390 | 52.1 | 1.2% | 1,386 | 45.9 | 3.3% |
| 2023 | 4,735 | 65.2 | 1.4% | 1,418 | 32.1 | 2.3% |
| 2024 | 4,856 | 86.6 | 1.8% | 1,487 | 35.1 | 2.4% |
| 2025 | 4,987 | 81.2 | 1.6% | 1,427 | 32.3 | 2.3% |
Sources: EIA movements of crude oil by rail (intra-U.S. and imports from Canada series); EIA Crude Oil Production, annual field production; EIA U.S. Crude Oil Imports, imports from Canada; all releases of October 30, 2026, checked November 4, 2026. The percentages are ratios calculated here, not shares of barrels shipped by rail: see the last section for why.
Read these as ratios. Domestic rail, in EIA's definition, includes barrels that move from one place to another inside the same PADD, production includes Alaska and the federal Gulf offshore, which together were 862.9 million barrels in 2025, 17.3 percent of the total (on the writer's reading, mostly volumes with no rail option), and a barrel can be counted twice if it rides a train to a pipeline and then moves on. The imports ratio has a different problem: the numerator is a movement estimated from waybills and the denominator is EIA's import statistics, so the two are not measured the same way. What the ratios do show is direction and scale. Even at the 2014 peak, domestic rail was equal to about one tenth of the production figure. Its weight was regional, in the places where rail was the marginal outlet. The difference between PADDs, and why the data are cut that way, is in PADD districts, explained.
The Bakken leg and the pipelines that replaced it
EIA's own articles supply the starting conditions. In March 2013 EIA reported that Bakken crude had sold as much as $28 a barrel below WTI in early 2012, that takeaway capacity out of the Williston Basin grew from about 678,000 barrels a day at the end of 2011 to over 1.1 million in 2012, and that rail was most of that growth: from an estimated 265,000 barrels a day in 2011 to approximately 660,000 in December 2012. EIA also said pipelines are "the most cost-effective way to transport crude oil in the United States, but they are expensive to build and may face regulatory hurdles", which it gives as a reason producers turned to rail. The inter-PADD tables show the result. Of the crude that left PADD 2 for another PADD in 2012, 41 percent went by rail; in 2013 the rail share peaked at 48.8 percent, and pipelines carried 39.2 percent.
Table 4. Crude oil leaving PADD 2 for PADDs 1, 3, 4 and 5, by mode, thousand barrels, with North Dakota production
| Year | Outbound total | By rail | By pipeline | By tanker and barge | Rail % | Pipeline % | North Dakota production | All rail from PADD 2, % of N.D. production |
|---|---|---|---|---|---|---|---|---|
| 2010 | 86,752 | 5,283 | 74,383 | 7,086 | 6.1% | 85.7% | 112,555 | 16.9% |
| 2012 | 240,195 | 98,473 | 116,894 | 24,828 | 41.0% | 48.7% | 242,354 | 49.2% |
| 2013 | 417,498 | 203,907 | 163,816 | 49,775 | 48.8% | 39.2% | 312,357 | 74.0% |
| 2014 | 539,357 | 239,950 | 265,914 | 33,493 | 44.5% | 49.3% | 393,562 | 66.3% |
| 2015 | 647,815 | 214,528 | 413,046 | 20,241 | 33.1% | 63.8% | 429,596 | 52.9% |
| 2016 | 572,455 | 122,948 | 436,511 | 12,996 | 21.5% | 76.3% | 380,393 | 32.8% |
| 2017 | 666,256 | 75,090 | 580,123 | 11,043 | 11.3% | 87.1% | 394,826 | 19.1% |
| 2018 | 795,740 | 84,068 | 686,346 | 25,326 | 10.6% | 86.3% | 466,395 | 18.1% |
| 2019 | 978,054 | 114,474 | 834,417 | 29,163 | 11.7% | 85.3% | 524,431 | 21.8% |
| 2021 | 748,001 | 59,680 | 670,051 | 18,270 | 8.0% | 89.6% | 409,104 | 14.6% |
| 2023 | 862,379 | 48,413 | 808,739 | 5,227 | 5.6% | 93.8% | 435,295 | 11.1% |
| 2025 | 858,293 | 52,436 | 800,666 | 5,191 | 6.1% | 93.3% | 427,561 | 12.3% |
Sources: EIA, Movements by Pipeline, Tanker and Barge, and Pipeline, Tanker, Barge and Rail between PAD Districts (crude oil columns for the pairs PADD 2 to PADD 1, 3, 4 and 5) and EIA movements of crude oil by rail; EIA Crude Oil Production, North Dakota; releases of October 30, 2026, checked November 4, 2026. For every pair and year the combined series equals pipeline plus tanker and barge plus rail. The last column is a ratio calculated here: it uses all rail from PADD 2, which includes rail from other PADD 2 states and moves inside PADD 2, so it is not the share of North Dakota crude that moved by rail.
The pipeline side of the table is where the story turns. Crude moving by pipeline from PADD 2 to the Gulf Coast was 49.6 million barrels in 2010, 750.9 million in 2019 and 693.8 million in 2025. EIA's September 2018 analysis names the projects behind the change: the Dakota Access Pipeline from the Bakken to Patoka, Illinois, and the Energy Transfer Crude Oil Pipeline from Patoka to Nederland, Texas, which raised the pipeline share of the Midwest's domestic crude shipments from 64 percent in June 2015 to 87 percent in June 2018. The annual figures in Table 4 land in the same place: 63.8 percent in 2015 and 86.3 percent in 2018. By 2025 pipelines carried 93.3 percent of what left PADD 2 and rail 6.1 percent.
The ratio in the last column of Table 4 gives a feel for Bakken-sized volumes without claiming more than it can. All rail from PADD 2 equalled 74 percent of North Dakota's production in 2013 and 66 percent in 2014, and 12 percent in 2025. The North Dakota Pipeline Authority publishes its own estimate of rail exports out of the state, and its chart of estimated rail volumes through August 2026 has the same outline as the EIA PADD 2 series: a steep rise through 2012 and 2013, a plateau in 2014 and early 2015, a decline into 2017, a partial recovery in 2018 and 2019, and a low plateau since 2021 with a modest rise in 2025 and 2026. The Authority adds a caution that applies to any estimate of this kind: "Rail estimate is a range due to pipeline flow assumptions and a delay in production data from Eastern Montana and South Dakota." The chart's shape is all that is taken from it; it carries no figures.
Table 5. All crude moved between PADDs, by mode, selected years, thousand barrels
| Year | All inter-PADD crude | By rail | By pipeline | By tanker and barge | Rail % | Pipeline % | PADD 2 to PADD 3 by pipeline |
|---|---|---|---|---|---|---|---|
| 2010 | 614,914 | 5,653 | 593,535 | 15,726 | 0.9% | 96.5% | 49,630 |
| 2012 | 693,351 | 103,624 | 561,100 | 28,627 | 14.9% | 80.9% | 81,095 |
| 2014 | 1,079,435 | 289,117 | 707,823 | 82,495 | 26.8% | 65.6% | 223,035 |
| 2016 | 1,163,939 | 137,728 | 1,000,827 | 25,384 | 11.8% | 86.0% | 347,163 |
| 2018 | 1,428,207 | 99,852 | 1,266,235 | 62,120 | 7.0% | 88.7% | 598,584 |
| 2019 | 1,665,243 | 125,016 | 1,484,246 | 55,981 | 7.5% | 89.1% | 750,908 |
| 2022 | 1,315,233 | 52,069 | 1,232,619 | 30,545 | 4.0% | 93.7% | 621,938 |
| 2025 | 1,535,903 | 77,812 | 1,418,686 | 39,405 | 5.1% | 92.4% | 693,829 |
Source: EIA, Movements by Pipeline, Tanker and Barge, and Pipeline, Tanker, Barge and Rail between PAD Districts, crude oil columns summed over all PADD pairs with a crude series, annual, release of October 30, 2026, checked November 4, 2026. Rail inside one PADD, and rail to or from Canada, is not in this table. Shares are calculated.
Across all PADD pairs the rail share peaked at 26.8 percent in 2014 and was 5.1 percent in 2025. Total inter-PADD crude grew from 615 million barrels in 2010 to 1,536 million in 2025, and pipelines supplied 825 million barrels of that 921 million increase. Pipelines carried 1,419 million barrels in 2025 against 708 million in 2014, while rail carried 78 million against 289 million. The Midwest's main storage and pricing hub is covered in the Cushing delivery-point explainer.
Where rail crude still goes
The routes that carry rail crude changed along with the volume. In 2014 the largest route was the Midwest to the East Coast, at 134.6 million barrels. In 2019 two routes were near equal: Canada to the Gulf Coast at 62.8 million barrels and the Midwest to the West Coast at 60.2 million. In 2025 the Midwest to the West Coast was larger than any other route, at 45.3 million barrels against 26.2 million for the next, 40 percent of all rail crude and 86 percent of everything that left PADD 2 by rail.
Table 6. Rail crude by route, thousand barrels, in 2014, 2019 and 2025, with each route's largest year
| Route | 2014 | 2019 | 2025 | Largest year and volume |
|---|---|---|---|---|
| PADD 2 to PADD 1 (Midwest to East Coast) | 134,571 | 42,771 | 6,858 | 2015: 135,169 |
| PADD 2 to PADD 3 (Midwest to Gulf Coast) | 52,447 | 11,459 | 322 | 2013: 88,757 |
| PADD 2 to PADD 5 (Midwest to West Coast) | 52,932 | 60,244 | 45,256 | 2019: 60,244 |
| PADD 4 to PADD 3 (Rocky Mountain to Gulf Coast) | 30,928 | 6,805 | 18,800 | 2014: 30,928 |
| PADD 4 to PADD 1 (Rocky Mountain to East Coast) | 9,402 | blank | 3,266 | 2014: 9,402 |
| Canada to PADD 3 (Gulf Coast) | 21,912 | 62,789 | 26,245 | 2019: 62,789 |
| Canada to PADD 1 (East Coast) | 26,360 | 16,669 | 1,810 | 2014: 26,360 |
| Canada to PADD 5 (West Coast) | 4,245 | 16,655 | 4,243 | 2019: 16,655 |
| Canada to PADD 2 (Midwest) | 637 | 14,096 | blank | 2021: 15,463 |
| All other routes, by difference | 48,600 | 19,736 | 6,705 | calculated |
| Total, all routes | 382,034 | 251,224 | 113,505 | 2014: 382,034 |
Source: EIA, Movements of Crude Oil by Rail, annual, release of October 30, 2026, checked November 4, 2026. 'Blank' means the EIA table has no value for that route and year. The all-other-routes row is the total less the nine listed routes, calculated here; it includes rail inside a single PADD and rail to Canada. PADD 1 is the East Coast, PADD 2 the Midwest, PADD 3 the Gulf Coast, PADD 4 the Rocky Mountain region and PADD 5 the West Coast.
The Midwest to West Coast line has no pipeline series in EIA's tables. EIA's list of pipeline movements between PADDs has no PADD 2 to PADD 5 pair, and the combined pipeline, tanker, barge and rail series for that pair equals the rail series in every year from 2010 to 2025. Within those tables, then, every barrel recorded moving from the Midwest to the West Coast went by rail. The volume was 0.5 million barrels in 2010, 52.9 million in 2014, a high of 60.2 million in 2019 and 45.3 million in 2025, so it fell far less than the East Coast line, which went from 135.2 million barrels at its 2015 peak to 6.9 million.
The Rocky Mountain region held up better than the Midwest. PADD 4 rail was 48.9 million barrels in 2014 and 28.7 million in 2025, with 18.8 million of the 2025 volume going to the Gulf Coast. EIA's tables do not say which fields feed the trains. Canadian rail kept one destination and lost most of the others: of the 110.2 million barrels from Canada in 2019, 62.8 million (57 percent) went to the Gulf Coast, 16.7 million to the East Coast, 16.7 million to the West Coast and 14.1 million to the Midwest. In 2025 the Gulf Coast took 26.2 million of 32.3 million, 81 percent, and the East Coast 1.8 million. The other end of the Gulf Coast's crude flows, the barrels that leave the country, is in US crude oil exports, explained.
East Coast refiners and the Brent-WTI spread
The East Coast case is the cleanest test of rail against price because the region has very limited pipeline access to domestic crude, so its choice is between rail, tanker and barge, and imports. EIA's September 2018 article says this directly: "Because very limited pipeline infrastructure exists to transport domestic crude oil to the East Coast (PADD 1), East Coast refiners historically have imported crude oil from international sources," and the Brent-WTI spread "helps determine when the additional costs of acquiring and transporting domestic crude oil to the East Coast can be overcome." Table 7 puts the East Coast's receipts from other PADDs beside the annual average Brent minus WTI spot price.
Table 7. East Coast (PADD 1) crude receipts from other PADDs by mode, thousand barrels, and the annual Brent minus WTI spot spread
| Year | Receipts from other PADDs | By rail | By tanker and barge | By pipeline | Rail % | Brent minus WTI, $ a barrel |
|---|---|---|---|---|---|---|
| 2010 | 16,282 | 783 | 6,780 | 8,719 | 4.8% | $0.13 |
| 2011 | 21,446 | 1,971 | 9,839 | 9,636 | 9.2% | $16.38 |
| 2012 | 29,441 | 20,045 | 4,330 | 5,066 | 68.1% | $17.58 |
| 2013 | 109,560 | 84,486 | 19,331 | 5,743 | 77.1% | $10.58 |
| 2014 | 192,233 | 143,973 | 42,951 | 5,309 | 74.9% | $5.80 |
| 2015 | 175,849 | 142,407 | 28,186 | 5,256 | 81.0% | $3.66 |
| 2016 | 78,803 | 59,217 | 14,665 | 4,921 | 75.1% | $0.35 |
| 2017 | 42,605 | 24,225 | 14,135 | 4,245 | 56.9% | $3.33 |
| 2018 | 68,563 | 28,473 | 34,866 | 5,224 | 41.5% | $6.11 |
| 2019 | 74,027 | 42,820 | 26,888 | 4,319 | 57.8% | $7.31 |
| 2020 | 67,795 | 33,752 | 30,013 | 4,030 | 49.8% | $2.80 |
| 2021 | 37,167 | 16,973 | 16,570 | 3,624 | 45.7% | $2.73 |
| 2022 | 42,469 | 12,475 | 26,517 | 3,477 | 29.4% | $6.03 |
| 2023 | 28,702 | 6,719 | 17,822 | 4,161 | 23.4% | $4.91 |
| 2024 | 37,416 | 8,666 | 25,375 | 3,375 | 23.2% | $3.89 |
| 2025 | 44,363 | 10,124 | 32,331 | 1,908 | 22.8% | $3.75 |
Sources: EIA, Movements by Pipeline, Tanker and Barge, and Pipeline, Tanker, Barge and Rail between PAD Districts (PADD 1 receipts from PADDs 2, 3, 4 and 5, crude oil columns) and EIA movements of crude oil by rail; EIA Spot Prices, Europe Brent (RBRTE) and Cushing WTI (RWTC), annual averages, release of October 28, 2026; all checked November 4, 2026. The table excludes imports and rail from Canada. Rail % is rail's share of the receipts column, calculated; the spread is the Europe Brent annual average minus the Cushing WTI annual average, calculated.
The East Coast's domestic receipts rose from 21.4 million barrels in 2011 to 192.2 million in 2014, and rail carried 68 percent of them in 2012, 77 percent in 2013 and 81 percent in 2015. The spread's annual average was $16.38 in 2011 and $17.58 in 2012, then $10.58, $5.80 and $3.66 in 2013 to 2015. Rail receipts climbed in the years after the spread opened and stayed high while it collapsed: East Coast rail receipts were 2.0 million barrels in 2011, 20.0 million in 2012, 84.5 million in 2013 and 144.0 million in 2014, a year in which the annual spread was $5.80. By 2025 rail carried 23 percent of the East Coast's 44.4 million barrels from other PADDs, and tanker and barge carried 73 percent.
EIA's 2018 article gives the clearest evidence that the spread alone does not decide the mode. Brent minus WTI averaged $5 a barrel in the first half of 2015 and again in the first half of 2018, yet domestic crude was 48 percent of East Coast refinery crude inputs in the first half of 2015 and 19 percent in the first half of 2018. The article attributes the change to new pipeline options out of the Midwest, to a fleet of coastwise-compliant tankers and barges built on the Gulf Coast that made shipping by water less expensive, and to East Coast refiners canceling or not renewing domestic supply contracts. The article puts Midwest rail shipments to the East Coast at 382,000 barrels a day in June 2015 and 75,000 in June 2018, and says that since mid-2017 the East Coast has received as much crude by tanker and barge from the Gulf Coast as by rail from the Midwest. On the annual figures, Gulf Coast crude arriving by tanker and barge was ahead of Midwest rail in 2018, 2022, 2023, 2024, 2025 and behind it in 2017, 2019, 2020, 2021; in 2018 the two were 31.0 and 27.9 million barrels. Annual averages in the table are not the half-year averages EIA quotes: the 2015 annual spread is $3.66 and 2018's is $6.11. The WTI side of that spread is explained in WTI vs Brent spread explained.
Canadian crude by rail, in two data sets
Two agencies measure the same flow from opposite ends. The Canada Energy Regulator (CER) publishes monthly Canadian crude exports by rail, compiled from the monthly reports that export-authorization holders file, "as the crude oil export crosses the Canadian border." EIA estimates what arrives, from U.S. waybills. The two series should be close and they are not identical. Table 8 sets them side by side, converting EIA's thousand barrels a year and CER's monthly barrels to thousand barrels a day.
Table 8. Canadian crude by rail: CER exports and EIA receipts from Canada, thousand barrels a day
| Year | CER exports by rail | EIA U.S. imports from Canada by rail | EIA compared with CER |
|---|---|---|---|
| 2012 | 46.3 | 21.3 | -54.0% |
| 2013 | 126.2 | 80.8 | -36.0% |
| 2014 | 161.0 | 145.6 | -9.5% |
| 2015 | 111.3 | 103.6 | -6.9% |
| 2016 | 87.9 | 90.5 | +3.0% |
| 2017 | 131.8 | 143.1 | +8.5% |
| 2018 | 242.2 | 238.4 | -1.6% |
| 2019 | 281.4 | 301.9 | +7.3% |
| 2020 | 169.7 | 169.2 | -0.3% |
| 2021 | 146.1 | 167.1 | +14.4% |
| 2022 | 143.5 | 125.8 | -12.4% |
| 2023 | 98.3 | 88.1 | -10.4% |
| 2024 | 88.7 | 95.9 | +8.2% |
| 2025 | 77.2 | 88.5 | +14.6% |
Sources: CER, Canadian Crude Oil Exports by Rail, Monthly Data, updated October 26, 2026 (annual figure is the sum of the twelve monthly barrel volumes divided by days in the year; the data start in 2012 'for confidentiality reasons'); EIA movements of crude oil by rail, imports from Canada series, release of October 30, 2026; checked November 4, 2026. The comparison column is calculated.
The gap ranges from EIA 54 percent below CER in 2012 to 15 percent above it in 2025, and changes sign several times, so neither source is a stable multiple of the other. Neither document explains the difference. EIA's methodology says it uses CER data to convert Canadian-origin carloads to barrels, and CER warns that its volumes are "subject to change" as monthly export reports are revised. The revision is visible: CER's December 2020 snapshot gave average rail exports of 231 thousand barrels a day for 2018 and over 273 thousand for 2019, while the monthly table updated October 26, 2026 sums to 242 and 281. Later years agree with CER's published snapshots: 98.3 thousand barrels a day for 2023, 88.7 for 2024 and 77.2 for 2025.
CER's April 2026 snapshot puts 2025 rail exports at its lowest since 2012 and gives the reasons: "Crude oil is increasingly being exported by pipeline or marine instead of rail, after additional pipeline capacity and dock loading capacity were brought online in recent years." Rail was 1.8 percent of Canadian crude exports in 2025, 2.1 percent in 2024, 2.5 percent in 2023 and 3.7 percent in 2022 by CER's accounting, and the pipelines took 92.3 percent in 2023. The capacity additions CER names are the Enbridge Mainline Line 3 Replacement, completed in October 2021 with 370 thousand barrels a day of capacity, and the Trans Mountain Expansion Project, in service in May 2024 with 590 thousand barrels a day. The peak month was February 2020, at 412 thousand barrels a day, and the Alberta production limits and pipeline ceilings behind the 2019 and 2020 rail highs are in Alberta production and takeaway limits. The price signal on the Canadian side is the Western Canadian Select discount, covered in WCS vs WTI explained, and the barrels landing in the U.S. refining system are counted in US oil imports by country.
What the spread had to cover
A common question about crude by rail is how wide a price spread had to be before rail paid. The published evidence is thinner than the question: a dated cost estimate for 2012, a threshold from the Canada Energy Regulator for 2020, and no current tariff. The statements below are laid out as published, and the conclusions are limited to what they support. The most useful arithmetic is the gap between rail and the alternative: where a pipeline existed, rail had to beat pipeline cost, not zero. The 2012 estimates put that gap at $5 to $10 a barrel ($10 to $15 for rail against $5 for pipeline).
Table 9. Dated statements on rail costs and the price spreads that justified rail
| Source and date | What it says | Figure |
|---|---|---|
| EIA, Today in Energy, July 26, 2012 | Rail costs an average $10 to $15 a barrel nationwide against $5 by pipeline (Wolfe Trahan estimates). Argus unit-train rates from the Bakken: about $12.75 to St. James, Louisiana, $12.25 to Port Arthur, Texas, around $15 to New York Harbor. | $10 to $15 rail; $5 pipeline |
| EIA, Today in Energy, March 19, 2013 | Bakken crude sold as much as $28 a barrel below WTI in early 2012 when takeaway was short. New rail capacity let it bypass Cushing and briefly sell at a premium to WTI. | up to $28 discount |
| EIA, Today in Energy, July 10, 2013 | The Bakken to Brent gap had narrowed to less than $5 a barrel in recent months. "The narrower spread reduces the incentive to ship oil to coastal refineries." | under $5 |
| EIA, Today in Energy, September 20, 2018 | Brent minus WTI averaged $5 a barrel in the first half of 2015 and again in the first half of 2018. Domestic crude was 48% of East Coast refinery crude inputs in the first half of 2015 and 19% in the first half of 2018. | $5 both times |
| CER, market snapshot, December 4, 2019 | Producers pay the transport cost to Cushing and the Gulf Coast. "Since transporting crude by rail is more expensive relative to pipelines, differentials must be wider to justify the use of rail and cover its higher cost." WCS-WTI averaged US$13.63 in 2016, US$12.68 in 2017, US$26.43 in 2018 and US$10.30 in January to April 2019 (price source named: Kent Group). | $10.30 to $26.43 |
| CER, market snapshot, December 2, 2020 | "However, differentials must widen to around US$20/bbl to justify crude-by-rail exports like earlier in 2020." WCS-WTI averaged US$26.86 in 2018, US$14.08 in 2019 and US$11.24 in January to September 2020 (price source named: Net Energy). Rail exports lag differentials by one to two months. | around $20 |
| CER, market snapshot, July 3, 2024 | The WTI-WCS differential widened from US$11.48 a barrel in June 2023 to US$25.27 in November 2023; rail exports rose from 74,900 to 127,700 barrels a day over the same months. | $11.48 to $25.27 |
| CER, market snapshot, April 8, 2026 | "When the WCS-WTI differential narrows, some crude-by-rail becomes an uneconomic option." The average WTI-WCS discount was US$13.84 in 2024 against US$17.85 in 2023. | $13.84 (2024) |
Sources: EIA Today in Energy articles of July 26, 2012 (id 7270), March 19, 2013 (id 10431), July 10, 2013 (id 12031) and September 20, 2018 (id 37073); Canada Energy Regulator market snapshots of December 4, 2019, December 2, 2020, July 3, 2024 and April 8, 2026; all checked November 4, 2026. The two CER snapshots of 2019 and 2020 print different 2018 averages because they name different price sources; each is quoted as printed. The 2012 rail and pipeline costs are estimates from third parties quoted by EIA, not tariffs.
Set the 2012 estimates against the Brent-WTI annual spreads in Table 7 and the result is mixed. EIA describes very limited pipeline infrastructure to the East Coast, so the comparison for it is with the whole rail rate, about $15 a barrel to New York Harbor in the 2012 Argus quote, not the gap over a pipeline. Only 2 of the 16 years had an annual average Brent-WTI spread of $15 or more: 2011 and 2012, at $16.38 and $17.58. The 2013 average was $10.58. Rail volume nearly doubled between 2012 and 2013 (152.0 to 296.1 million barrels) and was higher than in 2013 in both 2014 and 2015, when the annual spread was $5.80 and $3.66. After 2013 the annual average spread never reached $8 a barrel; the highest was $7.31, in 2019. So a single-spread test fits the 2011 and 2012 start of the build-out, fits 2013 only loosely (the $10.58 average sits below the $15 rate, yet volume nearly doubled that year) and does not explain the volumes of 2014 and 2015. The explanations the sources do offer are capacity and contracts. EIA's March 2013 article says rail represented most of the 2012 expansion in Williston Basin takeaway capacity, and CER's 2019 snapshot says that longer-term rail contracts "lowered rail's transportation costs" and that entering into them "creates a sunk cost, which entices shippers to continue using rail even if differentials narrow." The Brent-WTI average is also not the spread a Bakken barrel faced, which EIA's 2013 articles chart separately, and the two cannot be separated with these data.
The Canadian threshold, with its caveats
CER's 2020 statement is the only explicit threshold in the sources: "However, differentials must widen to around US$20/bbl to justify crude-by-rail exports like earlier in 2020." It is a statement about a particular moment, and it comes with the note that rail exports lag differentials by one to two months. CER's own figures show the relationship holding in some periods and not in others. Rail exports and the differential rose together in late 2018, when WCS-WTI averaged US$26.86 for the year by CER's 2020 snapshot. In 2019 the differential averaged US$14.08 yet rail exports hit an annual high, which CER ties to export pipelines that were full and to Alberta's production limits. And since mid-2022, CER says, rail exports "have not followed the WTI-WCS price differential" because of pipeline capacity and oil production volumes. The June to November 2023 move, from US$11.48 to US$25.27 with rail up from 74,900 to 127,700 barrels a day, is the cleaner recent case.
Two cautions apply to every dollar figure here. The 2012 estimates are fourteen years old. CER says rail's cost fell as shippers signed longer-term contracts and operators gained experience, which suggests the 2012 gap is too wide for the late 2010s; none of the cost sources gives a replacement figure. And the benchmark spreads are averages of a daily series, while a shipper decides cargo by cargo. For the pricing mechanics behind the spread, crude oil API gravity and sulfur explains why heavy and light crudes are priced apart, which is the quality half of any rail-versus-pipeline decision.
Rail in 2026 so far
The monthly series runs through August 2026, as released on October 30, 2026. January to August 2026 totals 88.3 million barrels, 19.8 percent above the same months of 2025 and above every January to August total from 2022 to 2025. It is 0.2 percent below 2021's 88.5 million. The growth is in U.S. origin rail: domestic rail is 28.7 percent higher, PADD 2 originations 32.5 percent higher and PADD 4 originations 22.2 percent higher, while rail from Canada is 4.1 percent lower. Domestic rail for the eight months, 69.1 million barrels, is above each of the January to August totals for 2021 to 2025 and below 2020's 71.3 million.
Table 10. Rail crude, January to August, selected years, thousand barrels
| Series | Jan to Aug 2021 | Jan to Aug 2024 | Jan to Aug 2025 | Jan to Aug 2026 | 2026 vs 2025 |
|---|---|---|---|---|---|
| Total | 88,485 | 80,712 | 73,693 | 88,306 | +19.8% |
| Domestic, intra-U.S. | 45,552 | 57,954 | 53,679 | 69,110 | +28.7% |
| From Canada | 42,932 | 22,758 | 20,015 | 19,195 | -4.1% |
| Originating in PADD 2 | 42,770 | 36,041 | 34,979 | 46,342 | +32.5% |
| Originating in PADD 4 | 2,341 | 21,913 | 18,621 | 22,764 | +22.2% |
Source: EIA, Movements of Crude Oil and Selected Products by Rail, monthly, thousand barrels, release of October 30, 2026 (data through August 2026), checked November 4, 2026. Sums of the eight monthly values; the percentage change is calculated. The most recent months can include model-based near-month estimates and can be revised.
The latest months deserve the caveat EIA attaches to them. The methodology says EIA may estimate one to three months from models when sample data are missing, and that waybills from smaller railroads can be incomplete in the first two months of a quarter. The rise could therefore be revised. What it does not do is change the larger picture: even at the 2026 pace of 363 thousand barrels a day, rail is about a third of its 2014 level. The 284 thousand barrels a day of domestic rail is 33 percent of the 2014 domestic average.
Tank cars and trains behind the barrels
Volumes in barrels hide how much equipment moves. EIA's 2012 and 2013 articles used about 700 barrels per carload and described unit trains carrying up to 85,000 barrels; the methodology's worked example, for a PADD 2 origin, gives 607 barrels per car. The methodology defines unit trains as 80 or more like cars. Table 11 converts the average daily volume at both conversion factors and counts the number of 80-car trains at 607 barrels a car (48,560 barrels a train). These are illustrations; EIA converts each quarter's cars by origin PADD, API gravity and weight, not with one factor.
Table 11. Illustration: carloads and 80-car trains per day equivalent to the average daily rail volume
| Period | Thousand barrels a day | Carloads a day at 607 bbl | Carloads a day at 700 bbl | 80-car trains a day at 607 bbl |
|---|---|---|---|---|
| 2013 annual average | 811 | 1,336 | 1,159 | 16.7 |
| 2014 annual average (peak year) | 1,047 | 1,724 | 1,495 | 21.6 |
| October 2014 (peak month) | 1,137 | 1,874 | 1,625 | 23.4 |
| 2019 annual average | 688 | 1,134 | 983 | 14.2 |
| 2023 annual average | 267 | 439 | 381 | 5.5 |
| 2025 annual average | 311 | 512 | 444 | 6.4 |
Sources: EIA movements of crude oil by rail, annual and monthly, release of October 30, 2026; EIA methodology Table 15 (607 barrels per car) and Today in Energy of July 26, 2012 and July 10, 2013 (700 barrels per carload), checked November 4, 2026. All columns are calculated; the October 2014 line divides the month's total by 31 days. Annual averages use 365 days.
At the October 2014 peak the volume equals about 1,874 carloads a day on the methodology's factor, or 1,625 on the 2013 article's, which is one tank car about every 46 seconds around the clock. The 2025 average of 311 thousand barrels a day equals 512 carloads a day, about 6 such trains a day, across the whole country. The two conversion factors differ by 15 percent, so any car count from this series is a range, not a number. Because the conversion depends on API gravity and carload weight, crudes of different gravity do not fill a car with the same barrels, which is why EIA converts by origin and quarter.
How far these numbers stretch
Five limits cover most of the ways this series gets misread. First, the table counts movements, not unique barrels. A barrel railed from the Bakken to a Midwest terminal and piped to a refinery sits in both the rail and the pipeline tables, so adding modes gives movements, not production. Second, PADDs are administrative regions: PADD 2 is not the Bakken, and PADD 4 is the Rocky Mountain region, so a route label names a pair of regions and not a field. Third, the production comparison pairs an EIA movement estimate with an EIA survey, and the denominator includes crude that has no rail option. Fourth, the barrel conversion changed in 2015 and the sampling rates changed in 2021, and recent months are provisional. Fifth, the EIA and CER series disagree by 54 percent in 2012, 36 percent in 2013 and by no more than 14.6 percent in any year from 2014 to 2025, so any single-year Canadian rail figure needs its source named.
The series also covers crude only. EIA reports ethanol, biodiesel, renewable diesel, asphalt, petroleum coke and hydrocarbon gas liquids in separate rail tables, and the methodology says natural gasoline rail movements are not published. Tank car safety rules, derailment history and rail-terminal capacity are outside the scope of these sources. The sources agree that pipelines are the cheaper mode, and the data fit that reading: EIA's July 2013 article described production outstripping pipeline capacity as rail grew, and the 2025 volumes are what rail carries once the pipelines are there, with the Midwest to West Coast route, which has no pipeline series in EIA's tables, the largest of what remains.
Crude oil by rail questions
How much crude oil moves by rail in the US?
EIA estimates 113.5 million barrels in 2025, an average of 311 thousand barrels a day, counting rail moves inside the United States, exports to Canada and imports from Canada. Of that, 81.2 million barrels moved within the United States, which equals 1.6 percent of the 4,987 million barrels of U.S. field production. From January to August 2026 the monthly series totals 88.3 million barrels, 19.8 percent more than the same months of 2025.
When did crude by rail peak in the United States?
2014 was the peak year at 382.0 million barrels, 1,047 thousand barrels a day. The peak month was October 2014 at 35.3 million barrels, about 1,137 thousand barrels a day. Rail from Canada peaked later, in 2019, at 110.2 million barrels. The 2025 total is 29.7 percent of the 2014 total.
What share of US crude oil production moves by rail?
Using EIA's intra-U.S. rail movements divided by U.S. field production, the ratio was 9.9 percent in 2014 and 1.6 percent in 2025. It is a ratio of two EIA series, not the share of oil shipped by rail: the rail series includes moves inside a single PADD, a barrel can move by rail and again by pipeline, and production includes Alaska and federal offshore barrels, most of which, on the writer's reading, have no rail option.
Why did crude by rail fall after 2014?
EIA's September 2018 analysis ties the fall in Midwest rail shipments to new pipelines, including the Dakota Access Pipeline from the Bakken to Patoka, Illinois, and the Energy Transfer Crude Oil Pipeline from Patoka to Nederland, Texas, and ties the drop in East Coast domestic crude receipts to a narrower Brent-WTI spread among other factors. In EIA's tables the pipeline share of crude leaving PADD 2 rose from 39.2 percent in 2013 to 93.3 percent in 2025, while the rail share fell from 48.8 percent to 6.1 percent. For Canadian crude, the Canada Energy Regulator points to Enbridge's Line 3 Replacement (October 2021) and the Trans Mountain Expansion (May 2024).
Where does crude oil by rail go today?
The largest single route in EIA's 2025 table is PADD 2 to PADD 5, the Midwest to the West Coast, at 45.3 million barrels, or 39.9 percent of the total. Next come Canada to the Gulf Coast (26.2 million) and the Rocky Mountain region to the Gulf Coast (18.8 million). In 2014 the largest route was the Midwest to the East Coast at 134.6 million barrels; in 2025 it was 6.9 million.
How many barrels does a rail tank car carry?
EIA's methodology works it out by quarter from the API gravity of the crude and the carload weight; its worked example, a 175,000 pound average carload at an API gravity of 40, gives 607 barrels per car. EIA's Today in Energy articles of 2012 and 2013 used about 700 barrels per car to convert Association of American Railroads carloads. At 607 barrels a car, the 2025 average of 311 thousand barrels a day is about 512 carloads a day.
How much more does it cost to ship crude by rail than by pipeline?
The cost sources give one dated figure: in July 2012 EIA cited estimates of $10 to $15 a barrel by rail and $5 by pipeline, with Argus unit-train rates of about $12.25 to $12.75 from the Bakken to the Gulf Coast and about $15 to New York Harbor. They carry no current tariff. The Canada Energy Regulator says "It generally costs less per barrel to ship by pipeline than by rail," and in December 2020 it wrote that differentials must widen to around US$20 a barrel to justify crude-by-rail exports from western Canada.
Is crude by rail rising in 2026?
EIA's monthly series shows 88.3 million barrels from January to August 2026 against 73.7 million in the same months of 2025, an increase of 19.8 percent. Domestic rail is up 28.7 percent and rail from Canada is down 4.1 percent. EIA's methodology says it may estimate one to three months of rail movements from models, and the series runs through August 2026, so the latest months can be revised.
Sources and check dates
Data and sources, all checked November 4, 2026: U.S. Energy Information Administration, Movements of Crude Oil and Selected Products by Rail, annual (release of October 30, 2026, next release August 31, 2027) and the monthly table (data through August 2026, next release November 30, 2026), with the definitions, sources and notes; U.S. energy-by-rail data methodology (March 2024; no newer edition found at that address) and EIA's crude-by-rail notice that the standalone page is no longer published. EIA movements between PADDs: pipeline, tanker, barge and rail, pipeline and tanker and barge (annual, release of October 30, 2026); Crude Oil Production; U.S. Crude Oil Imports by country (both release of October 30, 2026); Spot Prices (annual averages, release of October 28, 2026). EIA Today in Energy: July 26, 2012, March 19, 2013, July 10, 2013 and September 20, 2018. Canada Energy Regulator, Canadian Crude Oil Exports by Rail, Monthly Data (updated October 26, 2026) and market snapshots of December 4, 2019, December 2, 2020, July 3, 2024 and April 8, 2026. North Dakota Pipeline Authority, Rail Transportation (chart of estimated rail volumes through August 2026, used for its shape only). Table 9 quotes estimates by Wolfe Trahan and Argus Media as reported by EIA; neither was consulted directly. EIA's rail table covers movements within the 50 States and the District of Columbia and to and from Canada; rail to or from Mexico is outside its definition. The Canadian rail figures are for crude oil only.