Drunk driving deaths by county, and the problem with ranking states
We built a county-level view of alcohol-involved fatal crashes to put local numbers on our state pages. Doing it surfaced something that undermines most published rankings of this statistic, including the ones we were about to write.
States that test more, find more
The usual headline — «X% of fatal crashes in this state involved a drinking driver» — sounds like a measure of behavior. It is substantially a measure of police procedure. Whether a driver is recorded as drinking depends on whether anyone obtained a blood or breath sample, and states differ on that by a factor of eight.
Across all 51 jurisdictions — 50 states plus DC — the share of drivers tested and the share of crashes recorded as alcohol-involved move together with a correlation of 0.61. The two ends of the table make the point without any statistics at all.
Test the most drivers
Test the fewest
This does not mean the low-testing states have less drunk driving. It means nobody knows what they have, and their published figure is a floor rather than an estimate. A ranking that puts them at the bottom is rewarding them for not looking.
Where the most people die in these crashes
Absolute counts are the safer comparison: they do not depend on how many other fatal crashes a county had, though they still depend on testing.
| County | Alcohol-involved | All fatal crashes | Share |
|---|---|---|---|
| Harris County, TX | 467 | 1,524 | 31% |
| Los Angeles County, CA | 411 | 2,347 | 18% |
| Maricopa County, AZ | 321 | 1,648 | 20% |
| Riverside County, CA | 292 | 943 | 31% |
| San Bernardino County, CA | 279 | 1,133 | 25% |
| Dallas County, TX | 273 | 953 | 29% |
| San Diego County, CA | 227 | 840 | 27% |
| Clark County, NV | 192 | 644 | 30% |
| Bexar County, TX | 191 | 635 | 30% |
| Kern County, CA | 181 | 589 | 31% |
| Orange County, CA | 172 | 645 | 27% |
| Shelby County, TN | 160 | 684 | 23% |
| Hillsborough County, FL | 159 | 667 | 24% |
| Fresno County, CA | 149 | 512 | 29% |
| Sacramento County, CA | 149 | 557 | 27% |
| Duval County, FL | 143 | 494 | 29% |
| Wayne County, MI | 141 | 638 | 22% |
| San Joaquin County, CA | 131 | 367 | 36% |
| Cook County, IL | 128 | 1,123 | 11% |
| Orange County, FL | 126 | 566 | 22% |
Highest share, among counties with enough crashes to compare
Counties with fewer than 60 fatal crashes over three years are excluded: at that size one bad weekend moves the percentage by ten points.
| County | Share | Alcohol-involved | All fatal crashes |
|---|---|---|---|
| Douglas County, OR | 51% | 45 | 89 |
| Pueblo County, CO | 50% | 45 | 90 |
| York County, ME | 49% | 31 | 63 |
| Hawaii County, HI | 49% | 35 | 72 |
| Flathead County, MT | 48% | 29 | 60 |
| El Dorado County, CA | 48% | 33 | 69 |
| St. Charles County, MO | 47% | 43 | 92 |
| Colleton County, SC | 46% | 31 | 68 |
| Ector County, TX | 45% | 56 | 125 |
| Richland County, SC | 44% | 81 | 184 |
| Cumberland County, NJ | 44% | 30 | 69 |
| York County, SC | 43% | 43 | 101 |
| Ventura County, CA | 43% | 68 | 160 |
| Bastrop County, TX | 42% | 29 | 69 |
| Lexington County, SC | 42% | 56 | 135 |
| Yakima County, WA | 42% | 54 | 130 |
| Rock County, WI | 41% | 27 | 66 |
| Virginia Beach city, VA | 40% | 36 | 89 |
| Cuyahoga County, OH | 40% | 124 | 308 |
| Weld County, CO | 40% | 52 | 129 |
How this was built, and what it cannot tell you
Source is the NHTSA Fatality Analysis Reporting System, annual national files for 2020–2022, which are public domain. A crash counts as alcohol-involved when the vehicle file records a drinking driver. Counties come from the crash record’s own county code; the testing share is the proportion of drivers in fatal crashes with a usable alcohol test result in the person file.
Three limits worth stating plainly. This is a narrower measure than the federal 0.08 BAC standard, so every figure here understates. It covers fatal crashes only, which are a small and unrepresentative slice of all drunk driving. And it says nothing about cause: a county with a high share may have worse roads, longer distances to a hospital, or simply more diligent police.
We publish it because the county view did not exist anywhere we could find, and because the testing caveat is missing from every ranking we checked — including, initially, our own state pages, which we rewrote after running these numbers.
Reuse it freely. The underlying data is a US government work; our aggregation is yours to cite. If you want a cut we have not published, ask.