← The Crash Report
Existential Dread

Every Sedan Driver Drinks the Same. Only Some of Them Die.

Dale Impactor III • Clautomotive Desk
June 20, 2026
Scatter plot showing 22 sedan models with nearly identical impairment rates but death rates spanning two orders of magnitude

According to the toxicology reports, and there are a staggering number of them, impairment rates across 22 of the most popular sedans on American roads cluster so tightly around 20.1% that the standard deviation is 1.4 percentage points, a coefficient of variation of 7%, which is essentially statistical noise masquerading as variety.[1] One in five fatal-crash sedan drivers tests positive for alcohol or drugs regardless of whether they were piloting a Chevrolet Impala or a Tesla Model 3.

100×
Spread in death rates per 100M VMT across sedans with identical impairment rates

So the behavior is a constant, a flat line that barely registers a pulse. Twenty percent, give or take, from the Prius at 16.0% to the Charger at 22.7%, with every other sedan piled in between like commuters on the 405. But the death rates occupying the other axis of this chart look like someone sneezed on the spreadsheet: the Impala kills at 5.0 per 100 million VMT while the Model 3 sits at 0.05.[1] Same species of driver, same Friday night decisions, one hundred times the body count.

I ran this cross-tabulation expecting to find that bad cars attract bad behavior, that impairment rates would track fatality rates upward in some satisfying correlation proving that dangerous people buy dangerous cars, and the scatter plot came back looking like a heart-rate monitor flatlined on the x-axis while the y-axis had a seizure, which is not the finding anyone in the automotive insurance industry wants to see because it means the actuarial tables are pricing driver risk when they should be pricing vehicle risk.

The Fleet Age Problem

Before anyone emails: yes, the Impala's fatal fleet skews heavily toward 2000-2012 model years, with model year 2008 alone accounting for 401 of its 3,774 deaths, and every one of those vehicles predates the 2012 ESC mandate, lacks automatic emergency braking, and was crash-tested under protocols that did not evaluate the scenarios most likely to kill the people now driving them on roads shared with 5,000-pound electric SUVs.[2] The Model 3 is all 2017 and newer, which is not exactly an apples-to-apples comparison.

Fair enough, but a buyer shopping used sedans at $10,000 is not choosing between eras. They're choosing between an Impala and a Forte on the same lot, at the same price, with the same Carfax pop-up, and one of those cars is 100 times more likely to kill them per mile driven, and nobody puts that on the window sticker. Even filtering to 2015-and-newer model years, the Altima carries 1,185 deaths against the Model 3's 91, fleet-adjusted spreads that still span an order of magnitude without any corresponding difference in driver behavior.[1]

What This Breaks

Drunk driving campaigns assume the driver is the problem. Spend enough money on PSAs and ignition interlocks and the death rate drops. Except it won't, because 80% of sedan fatalities involve stone-sober drivers operating vehicles whose structural engineering determines survival probability with a precision that toxicology never could. The Ford Fusion at 1.23 deaths per 100M VMT and the Honda Accord at 3.07 have indistinguishable impairment profiles (19.4% vs 20.0%), yet the Accord kills at 2.5 times the rate.[1] Telling Accord drivers to stop drinking won't close that gap, but building a better car will.

Limitations

FARS captures fatal crashes exclusively, covering roughly 36,000 annual deaths out of approximately 6.7 million total crashes, so a vehicle with low fatality rates might still injure its occupants at alarming rates in survivable collisions we simply cannot see in this dataset. Death rates use VMT estimates rather than actual odometer readings, introducing approximately ±15% uncertainty for lower-volume models. Impairment rates reflect toxicology from fatal crashes only, and if a car's crash structure is good enough that only the most catastrophic impacts prove fatal, that selection bias could artificially concentrate impaired drivers among the dead. Fleet estimates do not account for regional concentration, urban versus rural driving mix, or the socioeconomic variables that correlate with both vehicle choice and crash exposure.

What to Do

If you're shopping for a used sedan under $15,000, ignore the marketing and check the FARS data. The Kia Forte (0.40 deaths per 100M VMT), Chevrolet Cruze (0.63), and Kia Optima (0.58) are dramatically safer than the Honda Civic (2.25), Nissan Altima (2.88), and Honda Accord (3.07) at comparable price points. Run your VIN at nhtsa.gov/recalls and look up your specific model and year at iihs.org/ratings. And if someone tries to sell you a 2008 Impala, understand that twenty percent of its fatal-crash drivers were impaired, exactly the same share as a Model 3, but a hundred times more of them ended up in the FARS database.

Sources & References

  1. NHTSA, Fatality Analysis Reporting System (FARS), 2014–2023. nhtsa.gov
  2. IIHS, “Life-saving benefits of ESC continue to accrue,” 2011. iihs.org
  3. NHTSA FARS Query Tool, crash data cross-tabulations. cdan.dot.gov
  4. IIHS, Vehicle Size and Weight. iihs.org

Source: NHTSA FARS 2014–2023. Impairment defined as BAC > 0 or drug-positive toxicology in fatal crashes. Death rates are estimated using VMT survey data and carry ±15% uncertainty for low-volume models. See methodology for caveats.