NHTSA FARS Crash Data Analysis for Expert Testimony

The Deposition That Turns on a Database

You're three hours into a deposition. The defense biomechanical expert has just testified that a 42 mph T-bone collision at a signalized intersection is 'consistent with minor soft-tissue injury at most.' Your client has two herniated discs, a torn rotator cuff, and a $1.4 million life care plan. The expert sounds credible. He cited his 22 years of experience.

You pull up a FARS query you ran the night before. In the five-year period from 2018 to 2022, NHTSA's Fatality Analysis Reporting System recorded 4,817 fatalities in side-impact collisions at intersections where the striking vehicle's travel speed was between 35 and 45 mph. You read him the number. Then you ask him to reconcile 'minor soft-tissue injury' with 4,817 dead people in functionally identical crash configurations.

That's NHTSA FARS crash data analysis doing what it's supposed to do. Not as background research. As a weapon.

What FARS Actually Contains

FARS is not a sample. It's a census. Every fatal motor vehicle crash on a U.S. public road since 1975 is in the database, coded across more than 100 data elements per crash. That's over 1.2 million fatal crashes as of the 2022 reporting year, the most recent complete dataset available from NHTSA's FARS portal.

Each record includes the crash environment (road type, speed limit, intersection geometry, lighting, weather), the vehicles involved (make, model, year, damage extent, rollover status, number of occupants), and the people (age, sex, seating position, restraint use, ejection status, injury severity on the KABCO scale). Since 2010, FARS also integrates data from state crash reports, death certificates, vehicle registration files, and EMS records.

For a PI attorney building a case theory, FARS answers a specific category of question: how often do crashes like this one kill people? And what distinguishes the ones that kill from the ones that don't?

You can query FARS directly through NHTSA's online tool, or download the raw SAS and CSV files for custom analysis. Most reconstructionists and biomechanical experts use the raw files. If you're working with a crash analysis platform built for litigation, FARS benchmarks can be cross-referenced against the physics of your specific collision.

Turning Census Data into Case Evidence

Raw FARS data isn't admissible by itself. A spreadsheet of 4,817 fatalities doesn't walk itself into evidence. You need to show the court why this population is comparable to your client's crash, and what the data distribution tells us about injury risk at the relevant Delta-V and impact direction.

Here's how that works in practice. Start with the PDOF. Your client was struck on the driver's side at roughly 3 o'clock. Filter FARS for side-impact crashes (IMPACT1 values corresponding to left-side or near-side principal direction of force). Then narrow by speed. FARS doesn't record Delta-V directly for most cases, but it does capture posted speed limits, travel speeds where available, and the manner of collision. If your crash reconstruction produces a Delta-V of 28 mph and a PDOF of 270 degrees, you can build a FARS cohort of near-side impacts at comparable closing speeds.

Now you have a reference population. Within that population, you can show fatality rates by vehicle type, occupant age, restraint use, and airbag deployment. You can show that in the cohort most similar to your client's crash, the fatality rate was 34%. You can show that among survivors in that cohort, the median injury severity was KABCO = 3 (incapacitating injury), not KABCO = 1 (possible injury).

This is where a tool like Silent Witness's free Delta-V calculator becomes useful in tandem with FARS analysis. You get a physics-based Delta-V estimate from your crash photos, then anchor it against the national fatality database to contextualize severity. Two independent data sources, one story.

'FARS is the only dataset that gives you a complete national picture of fatal crash outcomes. When a defense expert says an impact was trivial, FARS lets you show the jury how many people died in that exact scenario.'
Senior crash reconstructionist, 18 years in plaintiff-side forensic work

FARS and the MIST Defense

The Minor Impact Soft Tissue defense lives and dies on the claim that low property damage means low injury risk. Defense experts will point to minimal crush, intact bumper covers, and repair estimates under $3,000 to argue your client couldn't possibly need surgery.

FARS data complicates that narrative. In the 2022 dataset, 1,847 fatal crashes involved vehicles with 'minor' or 'functional' damage ratings on the FARS damage extent scale. People died in crashes where the car looked fine afterward.

You can't use FARS to prove your specific client's injuries, obviously. But you can use it to demolish the categorical claim that low damage equals low injury. When a defense IME doctor or biomechanical expert testifies that 'the forces involved were insufficient to cause the claimed injuries,' a FARS-derived fatality rate for comparable impacts turns that opinion into an outlier, not a consensus.

This matters especially in Daubert jurisdictions. Under Federal Rule of Evidence 702, expert testimony must be based on 'sufficient facts or data' and the product of 'reliable principles and methods.' An expert who ignores the largest crash fatality database in the world while opining on crash survivability has a methodology problem. Point it out in your Daubert motion. Point it out again at trial.

Limitations You Need to Know Before You Cite It

FARS only records fatal crashes. That's the biggest limitation and the one defense counsel will hammer. If you're citing FARS fatality rates to argue your client's crash was severe, expect the objection: 'This database only captures the worst outcomes. It tells us nothing about the thousands of similar crashes where nobody died.'

The objection has some merit. FARS doesn't capture the denominator of all crashes at a given speed and configuration. For that, you need the National Automotive Sampling System (NASS) or its successor, the Crash Report Sampling System (CRSS), which uses probability sampling to estimate the total national crash population. CRSS gives you the full severity spectrum. FARS gives you the fatal tail of the distribution.

The strongest approach uses both. CRSS tells you that approximately 22,000 near-side impacts occurred nationally at closing speeds between 35 and 45 mph in 2022. FARS tells you that 4,817 of those, or roughly 22%, were fatal. Together, they tell a story that neither dataset tells alone.

Other limitations to flag before trial: FARS relies on police crash reports, which sometimes miscategorize speed, restraint use, or crash type. The KABCO injury scale used in FARS is less precise than the AIS scale used in clinical settings. And FARS data typically lags by 12 to 18 months, so the freshest data available at any given trial may be two years old.

Getting FARS Data Into Evidence

FARS data enters testimony through an expert witness, not as a standalone exhibit. Your reconstructionist or biomechanical expert testifies to the analysis they performed using FARS, explains the methodology, and presents the results. The underlying data is a government record, which helps with authentication under FRE 902(5), but the analysis itself needs to satisfy Daubert or Frye depending on your jurisdiction.

Practically, this means your expert's report should include the specific FARS query parameters (years, crash types, speed ranges, vehicle classes, restraint status), the resulting dataset size, and the statistical outputs. Vague references to 'NHTSA data' won't survive a motion in limine. Specific query parameters will.

For deposition prep, run the FARS queries yourself before you hand them to your expert. Know the numbers cold. When you cross-examine a defense biomechanical expert who claims your client's 35 mph side impact was 'low severity,' you want the fatality count for that exact crash profile loaded and ready. Don't wait for redirect. Use it on cross.

Silent Witness reports already incorporate validated crash severity benchmarks aligned with federal data. If you're running your own FARS analysis alongside a Silent Witness biomechanical report, you get two layers of evidence grounded in the same physics. The Delta-V from the photos, the injury probability from the biomechanical model, and the population-level fatality context from FARS.

A Practical Workflow for Trial Prep

Here's what this looks like from intake to testimony. Your client comes in after a 40 mph rear-end on an interstate. You upload the crash photos. A Delta-V estimate comes back at 18 mph with a PDOF of 180 degrees. The biomechanical analysis flags AIS 2 cervical strain probability at 61% and AIS 3 disc herniation at 23%.

Now you open FARS. You query rear-impact crashes (PDOF 150 to 210 degrees) at travel speeds between 35 and 45 mph, model years 2015 and later, driver position, belted occupants. FARS returns 2,340 fatal crashes in that configuration over five years. You calculate the annualized fatality count. You cross-reference with CRSS for the estimated total crash population at those parameters.

Your demand package now includes three things the defense has to deal with: a physics-based Delta-V with a documented methodology, a biomechanical injury probability tied to that Delta-V, and a national fatality rate for the same crash profile. Each one is independently verifiable. Each one meets Daubert. Together, they're hard to dismiss.

The defense expert can still disagree. But disagreeing with validated physics, published biomechanical models, and a federal fatality census is a different posture than disagreeing with a treating physician's clinical impression. It puts the defense on the wrong side of the data.

If you want to see what your specific crash scenario produces before running the full FARS analysis, the free Delta-V calculator takes a few photos and about two minutes.

This content is for informational purposes and does not constitute legal or medical advice.

Frequently Asked Questions

Can FARS data be used in non-fatal injury cases?

Yes. While FARS only records fatal crashes, the fatality rate for a given crash configuration provides powerful context for injury severity arguments. If 22% of crashes matching your client's profile were fatal, arguing that the same profile produces only 'minor soft-tissue injury' becomes difficult. Pair FARS with CRSS data to capture the full severity spectrum.

How current is FARS data?

FARS data typically lags 12 to 18 months behind real time. The 2022 dataset, released in 2024, is the most recent complete year. Preliminary fatality counts for the current year are published by NHTSA quarterly, but lack the detailed crash-level coding needed for expert analysis.

Does FARS record Delta-V?

Not directly for most crashes. FARS captures posted speed limits, estimated travel speeds, and damage extent codes, but not physics-based Delta-V calculations. To use FARS alongside Delta-V analysis, you need a separate reconstruction (from EDR data, crash photos, or simulation) to establish the Delta-V, then query FARS for crashes at comparable speeds and configurations.

Will a judge allow FARS-based testimony under Daubert?

FARS is a federally maintained census database with documented methodology, peer-reviewed applications, and known error rates. It checks every Daubert factor. The key is presenting specific query parameters and a clear analytical framework rather than vague citations to 'NHTSA statistics.' Defense challenges typically target the expert's application of FARS, not the database itself.

What's the difference between FARS and CRSS?

FARS is a census of every fatal crash on U.S. public roads. CRSS is a probability-weighted sample of all police-reported crashes, fatal and non-fatal, designed to estimate national totals. FARS gives you exact fatality counts for specific crash profiles. CRSS gives you estimated total crash counts across all severity levels. Using both together provides the strongest evidence.