How this is made
An independent record of people killed walking or riding on Southern California's streets since 2020. It is not a government site. Its aim is that each person is counted and, where possible, remembered by name.
Who is counted
People killed by a motor vehicle while walking or riding, since Jan 1, 2020, in ten counties: San Luis Obispo, Santa Barbara, Ventura, Kern, Los Angeles, San Bernardino, Orange, Riverside, San Diego and Imperial.
- Walking includes using a wheelchair or another mobility device, and riding a skateboard or kick scooter without a motor (CHP records these as pedestrians).
- Riding means a bike, e-bike, motorized bike, e-scooter or e-skateboard.
- On a public road or in a parking lot, even if the driver struck the person on purpose.
- Not counted: deaths with no motor vehicle involved (a fall from a bike, a crash between two riders, a train), and people riding mopeds or motorcycles.
Where the information comes from
- Crash reports. The California Highway Patrol's California Crash Reporting System (opens in a new tab), checked each month. It gives the date, time and place, the vehicle, and whether the driver left the scene. It never gives names.
- Who they were. Names, ages, stories and case updates come from news coverage, coroner and police releases, obituaries and memorial pages, searched each month. Every detail is stored with the exact words of its source and a link, and reviewed by a person before it appears. A few people with no released name have a hometown or a line about who they were, when a source gave one; the same rules apply.
- BikinginLA. BikinginLA (opens in a new tab) has tracked bike deaths across Southern California for years, and its reporting inspired this dashboard. It is cited in most riders' records that have a researched source, more than any other outlet.
Where each person is on the map
About 40% of crash reports have no map coordinates, or clearly wrong ones. Those dots are placed at the intersection the report names, else at a street address, else at the center of the city; deaths known only from the news use the streets the article names. Each person's page says how their location was found, and the circle in the corner of the map lists people with no known location. People at the same point share one dot.
Checked against 300 reports that do have coordinates, half of these placements were within about 130 feet of the reported spot and 85% within about 800 feet. About 1 in 20 was at the wrong intersection.
Rates
The county and city tables show deaths per 100,000 residents per year, from U.S. Census Bureau estimates, using complete years only. Rates in italics rest on few deaths and can swing from year to year. Cities have at least 10,000 residents; one that many people drive through can look worse than its own streets are. Smaller places have no rate.
The bikeways column is miles of bike lane or path per 100 miles of street, from OpenStreetMap. It is a rough comparison, since bike lanes are mapped more completely in cities, and it shows where bikeways exist, not that they prevent deaths. The "By street and bikeway" chart likewise shows where people riding died, not whether bikeways work: it does not count how many people ride on each street, and some bikeways were built after the crashes.
Connecting news stories to crash reports
Crash reports carry no names, so a story is matched to its report by date, county, city, age, gender and place. The match is automatic only when one report clearly fits; close calls are checked by a person. A death known only from the news is marked "No official report yet" until its report appears, and "No official report found" once the crash is more than a year and a half old.
Why recent numbers are low
Police agencies can take months to send reports to the state, and about 1 in 20 arrives more than a year and a half after the crash. Recent counts keep rising, so they are shown as "at least", with those months lighter. A lower count this year does not mean the streets have become safer. Deaths of people riding also reach the news sooner, so recent months can show more riders than the final count will.
What the count can miss
The count is a minimum. A death appears only if it reached the state's crash records or the news. Missed: crashes on private property, deaths weeks after a crash that the report does not record, and reports that do not list the person as walking or riding.
Case updates, such as "no arrest yet", show the date a source last reported them and may have changed since. Some early details were entered by hand, with a link to their source, before the exact quote was stored for every detail.
Why fault findings are not shown
Each crash report records which person the investigating officer found most at fault, and whether a person walking was in a crosswalk. These are not published, here or in the downloads: they are usually made without the account of the person who was killed.
What is published, and what is not
- A person reviews every name, story and memorial link before it is published.
- Drivers are never named. Only what happened is reported: whether they stayed, and whether they were arrested, charged, convicted, or sentenced.
- A child is named only after their family or officials have made the name public.
- Photos are never copied here. Memorial and fundraising pages are linked instead.
- These are crashes, not "accidents".
The downloads
One row per person, as a CSV, a GeoJSON for maps, and a JSON with every detail and its source. Dates are written 2026-09-21, and mode is "pedestrian" (walking) or "bicyclist" (riding any of the five above). victim_vehicle says what someone was on: blank for walking or a bike, otherwise "wheelchair", "e-bike", "motorized bicycle" (a motorized bike), "e-scooter" or "e-skateboard". The GeoJSON has a few fewer people than the others: those with no known location. A blank field means unknown.
- status: "official" is in the state's crash records only; "confirmed" is in the crash records and has details from news or other sources; "reported" is known only from the news.
- precision: how the location was found. "reported" is the crash report's own coordinates; "intersection" and "address" were named in a report or news story; "near", "approximate", "city" and "county" are looser, as each person's page says; "researched" comes from news reports or the family.
- hit_and_run: true when the crash report or a news report says the driver did not stay; false when it says the driver stayed.
- dui_suspected: true when a source says the driver was suspected of driving under the influence, false when a source says not. It records a suspicion, not a finding, and it is blank unless a source says either, so blank does not mean no.
Corrections and removals
If you knew someone here and something is wrong, or you would like details added or removed, please write to orenbj@gmail.com. You will get a reply within a week. Removed details come off the site and out of the downloads; the death itself stays in the count, without a name.
Measuring visits
This site counts visits with Google Analytics. From this dashboard it also counts, without names or search words, that a person's page was opened and which data file was downloaded.