How to Pull Your Own Local Crash Data and Settle Whether a Road Is Really Dangerous

That intersection everyone calls a death trap? The crash record exists, it's public, and you can pull it free. Here's exactly where the data lives.

You've heard a neighbor swear the corner by the school is a death trap. You've felt your own stomach drop merging onto that on-ramp where nobody yields. Maybe someone you know got hit there. The feeling is real — but a feeling won't move a city council, and it won't tell you whether the road is genuinely dangerous or just unpleasant.

Here's the good news, and the core NU idea: the crash record already exists, it's public, and you can pull it yourself for free. Every reportable collision a police officer writes up becomes a row in a government database. You don't need a consultant or a records-request fee to see it. You need to know which database, and how to read it without fooling yourself.

Start national: NHTSA's FARS (fatal crashes, every state)

The single most trustworthy, nationwide source is the Fatality Analysis Reporting System (FARS), run by the National Highway Traffic Safety Administration. It covers every fatal crash on a public road in the U.S. going back to 1975, and it's free to query at the NHTSA FARS site (search "NHTSA FARS query").

The catch is in the name: fatal only. A road can be genuinely dangerous and have zero deaths — lots of injury crashes, lots of near-misses. So FARS is your floor, not your full picture. Use it to confirm the worst-case record, then go local for the rest. NHTSA also publishes the Crash Report Sampling System (CRSS), but that's a national sample for estimating trends — not the tool for "this exact intersection."

Go to the state crash system — the real workhorse

Below the federal level, every state keeps its own crash records database, usually run by the state patrol, DOT, or DMV. This is where injury and property-damage crashes live, which is most of what makes a road feel scary.

California is the example worth knowing because its tools are unusually good and totally free:

• SWITRS — the Statewide Integrated Traffic Records System, maintained by the California Highway Patrol. It's the official record, but the raw interface is clunky.
• TIMS — the Transportation Injury Mapping System, built by UC Berkeley's SafeTREC. TIMS takes the same SWITRS data and makes it mappable and queryable: draw a radius around an intersection, set a date range, and get every reported collision with severity, party type (pedestrian, cyclist, vehicle), and contributing factors. Free with a quick account.

Other states have their own versions — names vary, quality varies. Search "[your state] crash data portal" or "[your state] DOT traffic records." Many state DOTs now run public crash dashboards or interactive maps. If you can't find one, the state patrol's public-records page will point you to how the data is released.

Don't skip the city and county

Your local police department or county may publish crash data the state map doesn't surface cleanly. Search "[your city] police crash dashboard" or "[your city] open data crash." A lot of mid-size and large cities run open-data portals (often on Socrata or ArcGIS) with downloadable collision tables. If your town has adopted a Vision Zero plan, there's very often a public "high-injury network" map already built — the city has done the hardest part for you.

How to read it without fooling yourself

Pulling the numbers is the easy part. Reading them honestly is where most people — and a lot of loud Facebook posts — go wrong.

• Use a multi-year window. One bad year can be noise. Look at 3 to 5 years so a single freak crash doesn't define the road.
• Look at severity, not just counts. Ten fender-benders in a parking-lot entrance is a different problem than two crashes that put people in the hospital. Sort by injury severity.
• Find the pattern. Are crashes clustered at one turn? All left-turns? All after dark? All involving pedestrians? The type of crash tells you whether it's a design problem or just a busy spot.
• Normalize when you can. A road with more crashes may simply carry far more cars. If you can find traffic-volume counts (state DOTs publish "AADT" — annual average daily traffic), crashes-per-volume is far fairer than raw totals.
• Mind reporting gaps. Minor crashes go unreported, and databases lag a year or two for the most recent data. The record undercounts — it rarely overcounts.

The honest limit, and how to act

Public crash data tells you what happened and where. It does not by itself prove why, and it won't design a fix. That's the line: the record gives you standing, not a verdict on the cause.

But standing is exactly what you were missing. When you walk into a council meeting, an email to the city traffic engineer, or a neighborhood meeting with "five years of SWITRS shows nine injury collisions at this intersection, six of them left-turn, three involving pedestrians," you've stopped arguing about a feeling and started arguing about the record. That's harder to wave away — and it's the difference between "everybody says it's dangerous" and "here's the data, now fix it." Pull the record first. The feeling got you looking; the record is what gets you heard.