Off the Rails

Where the Rails Fail

Table 1 · By state

Every reportable accident since 1975, mapped to the state it happened in. The freight-heavy corridors light up: Texas and Illinois - the great classification-yard states - each carry more than 22k accidents on the record, an order of magnitude past the quiet New England end of the ramp. Color reads the raw count; the table below carries what color cannot - the hazmat cars that released and the people killed.

Shade encodes reportable accident count per state, five quantile classes. Alaska and Hawaii are inset by the Albers USA projection; territories fall outside its frame and appear only in the table.

Most accidents

01 Texas 21,922
02 Illinois 21,689
03 California 11,482
04 Pennsylvania 9,782
05 Ohio 9,032

Every state, in a table

State	Accidents	Hazmat cars released	Killed
Texas TX	21,922	497	372
Illinois IL	21,689	239	356
California CA	11,482	85	568
Pennsylvania PA	9,782	207	93
Ohio OH	9,032	184	122
Iowa IA	7,475	121	80
New York NY	7,379	77	101
Minnesota MN	6,814	53	53
Missouri MO	6,749	112	73
Indiana IN	6,720	86	192
Nebraska NE	6,385	61	85
Georgia GA	6,216	134	90
Kansas KS	6,182	74	90
Louisiana LA	6,040	166	124
Tennessee TN	5,413	90	50
Alabama AL	4,864	108	198
Wisconsin WI	4,826	54	46
Washington WA	4,732	32	63
Kentucky KY	4,689	111	33
Michigan MI	4,653	57	68
Arkansas AR	4,584	101	71
Virginia VA	3,889	42	70
Oklahoma OK	3,875	58	85
Florida FL	3,873	75	289
New Jersey NJ	3,378	28	63
Colorado CO	3,211	23	90
Oregon OR	3,211	27	23
Mississippi MS	3,208	68	126
North Carolina NC	3,047	87	82
Wyoming WY	2,958	45	45
West Virginia WV	2,876	58	30
Montana MT	2,746	78	37
Maryland MD	2,436	21	84
Idaho ID	1,848	24	29
North Dakota ND	1,843	57	56
Arizona AZ	1,830	104	41
South Carolina SC	1,817	42	94
Utah UT	1,776	16	15
Massachusetts MA	1,575	12	35
New Mexico NM	1,524	42	50
South Dakota SD	1,336	10	9
Connecticut CT	1,289	4	16
Maine ME	858	21	6
Nevada NV	855	14	27
Delaware DE	458	5	6
District of Columbia DC	446	3	0
Vermont VT	312	12	13
Alaska AK	284	60	6
New Hampshire NH	198	3	3
Rhode Island RI	119	0	5

One row per reporting state, summed across the whole 1975-2026 record. Accident count drives the map shade; hazmat-released cars and persons killed are the columns color cannot carry. See Methodology for how a "reportable" accident is defined.

The Named Roads

Table 2 · By railroad

Sixteen railroads account for the bulk of the file. The bar reads accidents; the columns beside it are the reckoning the annual report leaves out. Note the outlier: Amtrak reports almost no hazmat - it hauls people, not tank cars - yet leads on persons killed and injured per accident. Freight and passenger fail in different directions.

01
Union Pacific Railroad Company Class I
33,039 434 588 $4.0B
02
BNSF Railway Company Class I
17,037 329 328 $2.8B
03
CSX Transportation Class I
14,605 310 345 $1.5B
04
Burlington Northern Railroad Company Class I
13,414 155 214 $842M
05
Conrail Class III
12,897 203 165 $446M
06
Norfolk Southern Railway Company Class I
9,711 216 140 $925M
07
Southern Pacific Transportation Company Class I
8,957 189 194 $447M
08
Chicago And North Western Railway Company Class I
7,760 36 29 $209M
09
Amtrak (National Railroad Passenger Corporation) Not Assigned
5,830 0 773 $753M
10
Atchison, Topeka & Santa Fe Railway Company Class I
5,434 136 143 $339M
11
Illinois Central Gulf Railroad Company Class I
4,349 150 56 $168M
12
Louisville And Nashville Railroad Company Class I
4,154 102 54 $124M
13
Chicago, Milwaukee, St. Paul & Pacific Railroad Class I
4,145 14 21 $122M
14
Kansas City Southern Railway Company Class I
3,495 35 43 $274M
15
SOO Line Railroad Company Class I
3,485 43 27 $194M
16
Seaboard Coast Line Railroad Class I
3,345 70 54 $94M

accidents hazmat released killed damage

The Price of a Wreck

Table 3 · Cost & casualty per road

Divide each road's ledger by its accident count and the fleet sorts itself. The freight majors cluster low and to the right - expensive wrecks, few deaths per wreck. Then there is Amtrak, alone at the top of the chart: deaths per thousand accidents, roughly seven times any freight road. It hauls people, and when a passenger train wrecks, the cost is not measured in dollars.

One dot per railroad, 1975-2026 totals divided through by each road's accident count. Deep amber and a label mark Amtrak, the passenger outlier; the soft amber cluster is freight. Damage is nominal dollars, so long-dead roads read cheap. Hover any dot for its figures; the table carries all sixteen.

All sixteen roads, per-accident

Railroad	Accidents	Damage / accident	Killed / 1,000 accidents
Amtrak (National Railroad Passenger Corporation)	5,830	$129k	132.6
Atchison, Topeka & Santa Fe Railway Company	5,434	$62k	26.3
CSX Transportation	14,605	$102k	23.6
Southern Pacific Transportation Company	8,957	$50k	21.7
BNSF Railway Company	17,037	$166k	19.3
Union Pacific Railroad Company	33,039	$121k	17.8
Seaboard Coast Line Railroad	3,345	$28k	16.1
Burlington Northern Railroad Company	13,414	$63k	16
Norfolk Southern Railway Company	9,711	$95k	14.4
Louisville And Nashville Railroad Company	4,154	$30k	13
Illinois Central Gulf Railroad Company	4,349	$39k	12.9
Conrail	12,897	$35k	12.8
Kansas City Southern Railway Company	3,495	$78k	12.3
SOO Line Railroad Company	3,485	$56k	7.7
Chicago, Milwaukee, St. Paul & Pacific Railroad	4,145	$29k	5.1
Chicago And North Western Railway Company	7,760	$27k	3.7

Sorted by deaths per thousand accidents. Derived at build time from the same per-road totals as Table 2.

The Long Fall

Table 4 · By year

The record opens at the bottom of the industry's worst decade: rotten ties, bankrupt roads, and 13,729 wrecks in 1978 alone. Then the Staggers Act let the railroads shed their worst track, the FRA tightened inspection rules, and the count fell off a cliff - to a plateau near 2,332 a year, where it has sat for three decades. The fall is history; the plateau is the present tense.

Each column is one reporting year. The tallest is 1978 (13,729); 2026 is a partial year, reported through spring. Amber marks the peak. Hover any column for its exact count.

The Rare Catastrophe

Table 5 · Hazmat released, by year

Across 224,704 accidents, only 3,688 cars have ever actually released their cargo - a fraction of one percent. That is the whole shape of hazmat risk on rail: almost never, and then Graniteville, then East Palestine. The curve fell with the accident count; what did not fall is the size of a bad day.

Hazmat cars that released, per reporting year. The 1978 peak sits in the same era as the accident peak; the 2023 dot is the year of the East Palestine derailment (36 cars released nationwide that year). The hollow dot is 2026, a partial reporting year. A released car counts once, whether it held chlorine or corn syrup.

What Kind of Wreck

Table 6 · By accident type

One category swallows the file. Derailments are 64% of all reportable accidents - and 89% of every hazmat car that has ever released. Whatever a tank car is carrying, the way it gets out is almost always the same: the wheels leave the rail. Collisions are the rarer, deadlier cousins; spectacular categories like explosion barely register.

Derailment 143,717 · 64% 3,275 spilled
Side collision 21,977 · 10% 91 spilled
Other impacts 21,953 · 10% 113 spilled
Hwy-rail crossing 11,909 · 5% 63 spilled
Other (describe in narrative) 4,977 · 2% 12 spilled
Rear end collision 4,823 · 2% 45 spilled
Raking collision 4,471 · 2% 18 spilled
Obstruction 4,410 · 2% 17 spilled
Fire/violent rupture 3,185 · 1% 15 spilled
Head on collision 2,359 · 1% 18 spilled
Broken train collision 747 · 0% 17 spilled
RR grade crossing 89 · 0% -
Explosion-detonation 87 · 0% 4 spilled

Bar length is accident count (shared scale). The right figure is hazmat cars that released within that accident type - concentrated almost entirely in derailments. Counts sum across 1975-2026.

Why They Derail

Table 7 · By primary cause

Every Form 54 names a primary cause, and the causes are unglamorous. The single biggest killer of trains is wide gauge - rails spreading apart over rotten crossties until the wheels drop between them. Next come the yard's human failures: a switch lined the wrong way, a shove with nobody riding the leading end. Nothing on this list is exotic; that is the point. The family tag on each row says which kind of neglect it was.

Wide gage (due to defective or missing crossties) Track
12,303
Switch improperly lined Human factor
9,027
Shoving movement, absence of man on or at leading end of movement Human factor
7,621
Switch point worn or broken Track
7,171
Buffing or slack action excessive, train handling Human factor
5,946
Cross level of track irregular (at joints) Track
5,032
Other causes (Provide detailed description in narrative) Miscellaneous
4,870
Other miscellaneous causes (Provide detailed description in narrative) Miscellaneous
4,755
Shoving movement, man on or at leading end of movement, failure to control Human factor
4,135

Top nine primary causes across 1975-2026, each labeled with its FRA cause family. Bar shade distinguishes the family; the text tag carries it for readers who cannot rely on color.

The Blame Shifts

Table 8 · Cause family, by era

Rebuild the track and the track stops failing - mostly. In the deferred-maintenance 1970s, 37% of accidents were laid to track defects. Two decades of rebuilt roadbed cut that share by a quarter, and around the turn of the century the human factor overtook it. The modern era wrecks less - but when it wrecks, it is likelier than ever that a person, not a rail, failed first.

1975-84 93,295

1985-94 35,363

1995-04 36,257

2005-14 31,299

2015-26 28,490

Each bar is one era's accidents, split by the FRA's five cause families; the right figure is the era's total. Shades stay fixed per family across this page. Counted from the Form 54 cause-code column, binned by year.

Every era, every family, in a table

Era	Track	Human factor	Equipment	Signal	Miscellaneous	Total
1975-84	34,779 (37%)	29,489 (32%)	16,363 (18%)	193 (0%)	12,471 (13%)	93,295
1985-94	10,990 (31%)	13,097 (37%)	4,771 (13%)	276 (1%)	6,229 (18%)	35,363
1995-04	10,854 (30%)	14,135 (39%)	3,852 (11%)	675 (2%)	6,741 (19%)	36,257
2005-14	8,786 (28%)	12,009 (38%)	3,215 (10%)	735 (2%)	6,554 (21%)	31,299
2015-26	5,918 (21%)	11,257 (40%)	3,085 (11%)	699 (2%)	7,531 (26%)	28,490

Accident counts by primary-cause family per era, with each family's share of the era in parentheses. The family is read off the first letter of the FRA cause code (T / H / E / S, all else miscellaneous).

Slow Yards, Fast Mains

Table 9 · By track type

Half of everything in this file happens in a yard, at a walking-pace 4.5 mph average - cheap, unglamorous derailments over a mislined switch. But follow the money and it runs the other way: the main line, where trains average 22 mph, takes 71 cents of every damage dollar and 95% of the deaths. The yard wrecks often; the main line wrecks big.

Share of accidents of 224,704 wrecks

Share of damage of $17.6B reported

Yard: 4.5 mph avg
Main: 22 mph avg
Industry: 4.9 mph avg
Siding: 7 mph avg

Both panels rank the same four track types, so a row reads straight across: yards lead on wrecks, main line leads on damage. The strip below is the mean reported train speed at the moment of the accident - the reason the money and the deaths ride the main line. Counted from the Form 54 track-type column.

Speed Is the Multiplier

Table 10 · By train speed

Most trains wreck at a crawl: 167,276 accidents - the bulk of the file - happened at 10 mph or less, and almost everyone walked away. Push past 60 mph and the count collapses to 3,205, but each wreck is about 45 times as lethal. The left panel is the yard's story; the right panel is physics.

Accidents by speed at the moment of the wreck

Killed per 10,000 accidents same bands, same order

Same five speed bands, same order, two readings: how often that band wrecks, and how deadly the wreck is when it comes. Deep amber marks each panel's extreme - the crawl-speed pile-up on the left, the over-60 lethality on the right. Counted from the Form 54 train-speed column.

A Season for Failure

Table 11 · By month

Steel keeps a calendar. In deep cold it contracts until rails snap and joints pull apart - Mar is the worst month on the record. In July heat it does the opposite, buckling into sun kinks that throw trains at speed. The quiet months are the mild ones: Sep is the trough. The railroad's enemy is not weather in general; it is the thermometer's extremes.

Reportable accidents by calendar month, summed across 1975-2026. Deep amber marks the winter months, when broken rails and pull-aparts spike; the July shoulder is heat-kink season. Peak and trough are labeled; hover any column for its count. Counted from the Form 54 accident-month column.

All twelve months, in a table

Month	Accidents	Share of year
Jan *	19,639	8.7%
Feb *	18,612	8.3%
Mar	19,981	8.9%
Apr	18,260	8.1%
May	18,348	8.2%
Jun	18,804	8.4%
Jul	19,617	8.7%
Aug	18,989	8.5%
Sep	17,539	7.8%
Oct	18,483	8.2%
Nov	18,014	8.0%
Dec *	18,418	8.2%

* winter months (Dec-Feb), the emphasized columns above.

Two roads, side by side

Put any two railroads head to head - accidents, hazmat released, deaths, damage - every bar on one shared scale, so the comparison is honest by construction.

Open the compare tool →

Off the
Rails