Most Dangerous Roads for Pedestrians and Cyclists in the US (Part 2)

This is a follow up post to an analysis I worked through on October 17th about the most risky road segments in the U.S. for driving and generally. The following day, Oct. 18, 2026, The Wall Street Journal published an analysis of 1.5 miles of Union Avenue in Memphis, Tennessee by Rebecca Cadenhead and Andrew Mollica. The Journal counted 10 pedestrians killed on that stretch since 2020. Union Avenue has six lanes and carries about 30,000 cars a day.

The Journal used Union Avenue to show how people on foot are put at risk on a common American road design, the “stroad”. A stroad is a road that serves shops and homes, like a street, while also moving fast traffic, like a highway.

We ran the Journal’s test on every arterial road in the United States. The stretch with the most pedestrians killed is 1.5 miles of Westheimer Road in Houston. From 2020 to 2024, 19 people walking were killed there. A road with the same traffic would have about one pedestrian death in five years.

Union Avenue ranks 137th. It is one example of a road design that repeats in cities across the country.

Wherobots ran this analysis on public federal data. Wherobots is the AI Context Engine for Physical World Data: a platform that joins location data, such as crash records, road maps and traffic counts, at planetary scale.

How we extended the Journal’s analysis

The Journal looked at one street with Tennessee’s crash records. We looked at all of them with the national record. We kept the Journal’s rules where the national data allowed it, and added a test for chance.

1. Deaths. We took every pedestrian and cyclist killed from 2020 to 2024 from the Fatality Analysis Reporting System (FARS), the national record of every fatal traffic crash. That is 36,075 pedestrians and 5,247 cyclists.
2. Roads. We took every principal arterial, minor arterial and major collector from the road inventory that states report to the Federal Highway Administration. These are the busy roads between freeways and side streets, about 918,000 miles in all. Freeways are left out, as in the Journal’s analysis.
3. Stretches. We joined road pieces into continuous streets and slid a 1.5-mile window along each one, 0.1 mile at a time. A death counts for a window when the crash record places it within 200 feet of the road, the Journal’s rule.
4. Chance. On a 1.5-mile stretch, two or three deaths in five years can come from chance. We used the Empirical Bayes method from the Highway Safety Manual, the reference U.S. road agencies use to find dangerous roads. With it, we weigh each stretch’s deaths against the deaths on roads with the same traffic. A stretch ranks high only when its count is both large and far above that norm.
5. Checks. We picked the top 25 so that no two share a death. We checked that each one is a single, continuous piece of road 1.5 miles long. One stretch in Phoenix failed that check, and we set it aside. We also checked that the street names on each crash record match the road.

We screened every arterial and major collector outside freeways. Of the 36,075 pedestrians killed, 25,433 (70%) died within 200 feet of those roads and are counted. Another 1,315 died within 200 feet of them in crashes the federal record places on a freeway, so they are not counted. Our count of pedestrians killed in 2021, 7,470, matches the figure in the National Highway Traffic Safety Administration’s pedestrian fact sheet.

The national record has two limits the Journal’s did not. It holds deaths only, so injury crashes are not counted. It also ends in 2024, so deaths in 2025 and 2026 are not counted.

The 25 most dangerous roads for pedestrians

The map shows the 1.5-mile stretches with the most pedestrians killed from 2020 to 2024, after the Highway Safety Manual adjustment. Click a stretch for its rank, the deaths counted and the number expected. “Expected” is the number of pedestrian deaths a road with the same traffic would have in five years.

Central Avenue in Albuquerque had 17 deaths, the second most of any stretch. It ranks seventh because its traffic is lower, and under the method we give a low-traffic road’s count a little less weight. With a narrower 100-foot rule, all of the top 10 stay in the top 25. With shorter 0.5-mile stretches, 8 of the top 10 stay.

Each stretch’s records are in the full map. Choose a stretch to list every person killed there, with the date, the time and the nearest cross street. Each row also says whether a crosswalk was present and gives the case number in FARS, the federal record of fatal crashes. A reporter can request the police report with that case number. The list downloads as a spreadsheet file (CSV).

Where Union Avenue ranks

Union Avenue’s worst 1.5 miles runs from Cleveland Street east to Florence Street. In the national record, 7 pedestrians were killed there from 2020 to 2024. A road with its traffic would have about 0.5 deaths in five years, so Union Avenue had about 15 times the expected number. The Journal counted 10 deaths on its stretch, from Tennessee’s records through September 2026.

Union Avenue ranks 137th among U.S. stretches. We scored about 2.3 million windows on roads with at least one pedestrian or cyclist death, and 136 stretches had a worse record than Union Avenue.

Bakersfield, California, also has a Union Avenue in the top 25. Its 1.5 miles at number 13 had 13 pedestrians killed, all of them after dark.

What the most dangerous stretches have in common

The 25 stretches look like Union Avenue. Every one has at least four through lanes, and 17 have about six or more. The middle stretch carries about 52,000 vehicles a day.

Most deaths happened at night. On the 25 stretches, 220 of the 266 pedestrians killed died in the dark, or 83%. The Journal reported the same pattern: people crossing between intersections at night are the most likely to die. The vehicle struck the pedestrian first in 254 of the 266 deaths.

Florida has six of the 25 stretches, California five and Texas three. Two stretches are on U.S. 441 in Broward County, Florida, and two are on Beach Boulevard in Stanton, California. The stretches in each pair are less than 3 miles apart and share no death.

The most dangerous roads for cyclists

Cyclist deaths are rarer, so 1.5 miles holds too few to compare. No 1.5-mile stretch had more than 5 cyclists killed in five years. We used 5-mile corridors instead. We ranked them by how many more cyclists were killed than the number expected for the corridor’s traffic. Click a corridor on the map for its rank, the cyclists killed and the number expected.

Florida has 19 of the 25 corridors with the most cyclists killed above the expected number. The other six are in Arizona (2), California, Hawaii, Louisiana and Nevada. The top corridor, U.S. 19 in Pasco County, is the same road that ranks first in our ranking of the most dangerous highway stretches.

The counts are small: each of the 25 corridors had 3 to 7 cyclists killed. Ranked by raw count instead, all of the top 10 still appear in the top 25.

The cities where walking and biking are most dangerous

We also counted deaths in each of the 327 U.S. cities with 100,000 or more residents, on all roads. Where a city has few deaths, we adjusted its rate toward the average of all 327, the same Empirical Bayes idea used for the roads.

Memphis has the highest rate of pedestrian deaths. From 2020 to 2024, 304 people walking were killed in the city, about 9.8 a year per 100,000 residents. The next four cities are in the South:

• Jackson, Mississippi: 70 killed, 8.7 a year per 100,000 residents
• Fort Lauderdale, Florida: 84 killed, 8.6
• Little Rock, Arkansas: 88 killed, 8.5
• Baton Rouge, Louisiana: 90 killed, 8.0

For cyclists, Fort Lauderdale has the highest rate, with 17 killed, about 1.3 a year per 100,000 residents. New Orleans, St. Petersburg, Tampa and Baton Rouge follow, each about 1.1.

Cities differ in how many people ride a bike, so we also compared deaths with the number of people who bike to work. That count comes from the Census Bureau’s American Community Survey. Among cities with at least 500 bike commuters, Baton Rouge ranks first, with about 40 cyclists killed a year per 10,000 bike commuters. Jacksonville, Fresno, Detroit and Orlando each have about 35 to 38. New York has about 3 and Chicago about 4.

Data and rules

• Deaths: FARS 2020 to 2024, the person file. Pedestrians and cyclists killed in a crash with a vehicle on a public road.
• Roads and traffic: the Highway Performance Monitoring System, the road inventory states report to the Federal Highway Administration, 2024 edition.
• Residents and commuters: American Community Survey estimates for 2020 to 2024, inside Census place boundaries.
• Method: the Federal Highway Administration’s guidebook for finding high pedestrian crash locations, with the network screening method in the Highway Safety Manual.
• Expected deaths come from a model of deaths against vehicle traffic on each type of road. They are modeled, not counted.
• Computation: every join, count and model ran in Wherobots. The pipeline and the method checks are public in our working notes.

What the ranking leaves out

No national count of people walking or biking exists. The expected numbers use vehicle traffic only, so a busy sidewalk and an empty one look the same to the model. Deaths are a floor: crashes the record places more than 200 feet from the road are not counted, and neither are injuries.

About 30% of pedestrian deaths are not counted. The federal record places 6,123 of them on a freeway, which the ranking leaves out. The other 4,519 lie more than 200 feet from the roads we screened or have no recorded location.

Why AI is blind to the most dangerous roads

Every number in this post comes from joining three kinds of records: crash locations, road lines and population counts. AI models are trained on text, documents, databases and the internet. They can describe a stroad. Finding the worst ones takes a join of about 41,000 crash records against 918,000 miles of road, and Wherobots runs that join in one engine.