Two metrics, two different stories about road risk

The national average road fatality rate in 2023 was 12.21 deaths per 100,000 people, according to NHTSA FARS. By a different measure, 1.26 deaths occurred for every 100 million vehicle-miles driven. These two numbers describe the same underlying toll, 40,901 deaths, but from fundamentally different perspectives. One asks: out of every 100,000 people living in a place, how many die on the road? The other asks: out of every 100 million miles actually driven, how many drivers, passengers, and pedestrians die? The answer to the first question depends heavily on how much a state's residents drive. The answer to the second isolates the per-mile danger of roads and vehicles in a state, regardless of how many people choose to drive long distances.

For most densely populated states, the two rankings align reasonably well. For rural, low-population states in the interior West and rural South, the rankings diverge in ways that matter for policy and for how residents understand their own risk. A state can rank as extremely dangerous on a per-capita basis simply because its residents drive enormous distances, not because its roads or drivers are especially dangerous per mile traveled. The inverse is also true: a state with modest per-capita deaths might still have per-mile rates well above average if its crashes are concentrated on a small, heavily used road network.

Per-capita top 8: rural dominance

The eight states with the highest fatality rates per 100,000 population in 2023 are all either rural, sparsely populated, or both: Mississippi (24.9 per 100k), Wyoming (24.7 per 100k), New Mexico (20.7 per 100k), South Carolina (19.5 per 100k), Arkansas (19.4 per 100k), Alabama (19.1 per 100k), Tennessee (18.6 per 100k), Montana (18.4 per 100k). Mississippi leads at 24.9 per 100,000, more than double the national 12.21 average. Wyoming sits at 24.7, nearly identical to Mississippi on this metric.

What these states share is low population density combined with a driving culture of necessity. Residents of Wyoming or Montana routinely travel 40, 80, or 100 miles for medical care, employment, or routine shopping, trips that simply do not occur for residents of Boston or Manhattan. When every resident drives far more than the national average, the per-capita death count rises not because each mile is more dangerous but because each person accumulates far more miles of exposure. A per-capita ranking conflates high exposure with high danger, and the two are not the same thing.

Per-VMT top 8: a partially different list

Sorting by fatalities per 100 million vehicle-miles produces a partially different top 8: Mississippi (1.8 per 100M VMT), Arizona (1.7 per 100M VMT), South Carolina (1.7 per 100M VMT), Kentucky (1.7 per 100M VMT), West Virginia (1.6 per 100M VMT), Oregon (1.6 per 100M VMT), Tennessee (1.6 per 100M VMT), Oklahoma (1.6 per 100M VMT). Mississippi still leads, now at 1.8 per 100M VMT, meaning it is dangerous by both measures. Arizona ranks near the top of the per-VMT list despite a per-capita rate of only 17.5, which ranks it lower on the per-capita table. Arizona's road network includes high-speed arterials, Sun Belt pedestrian fatalities, and a large tourism VMT base, all contributing to its per-mile danger.

Wyoming, which ranks second on the per-capita list, drops in the per-VMT ranking. Its residents drive so many miles that when fatalities are divided by those miles, the per-mile risk normalizes relative to shorter-driving states. This is the clearest illustration of the metric's effect: Wyoming is dangerous to live in (high per-capita rate) because everyone drives far, but once you account for how much driving is happening, each individual mile is not as dangerous as Arizona's or Mississippi's miles.

Mississippi: dangerous by every measure

Mississippi is the clearest case: it ranks first on both metrics. At 1.8 per 100M VMT and 24.9 per 100,000 residents, Mississippi's roads are dangerous both in absolute per-person terms and in per-mile-driven terms. Neither high rural driving exposure nor a specific road engineering anomaly explains this alone. The state has documented challenges across multiple contributing factors: among the lowest seatbelt-use rates in the nation, high rates of impaired driving, a road network with significant proportions of undivided two-lane rural highways, and limited trauma care in rural areas. The per-VMT figure means that even after controlling for how much Mississippians drive, each mile is still far more likely to end in a fatality than the national average.

Massachusetts and the safe-state cluster

At the other end of both rankings, Massachusetts records 4.9 deaths per 100,000 residents and 0.6 per 100M VMT, the lowest rates in the country on both measures. Massachusetts benefits from high urban density (shorter trips, lower-speed crashes), robust public transit reducing vehicle exposure, relatively strong distracted-driving and seatbelt enforcement, and proximity to well-resourced Level 1 trauma centers. New York and the District of Columbia also rank near the safest end of both lists for similar structural reasons. These states demonstrate that low per-capita AND low per-VMT rates are achievable simultaneously, primarily through urban density and multimodal transportation infrastructure.

Which metric should you use?

For comparing individual risk to a resident, the per-capita rate is the more intuitive number: it answers the question "how likely am I to die on the road if I live here?" For comparing the engineering quality, road design, speed environment, and enforcement effectiveness of a state's road network independent of how much people happen to drive, the per-VMT rate is more useful. Both metrics have legitimate applications, and both are published in our state profiles. When a journalist or researcher uses only per-capita rates to rank state road safety, they are measuring the combination of driving culture and road danger, not road danger alone. When they use only per-VMT rates, they may miss the practical safety experience of residents who have no alternative to long daily drives. Neither framing is wrong; both are incomplete. The full picture requires both.

Policy implications follow directly from this framing. If a low-population rural state scores poorly on per-capita but adequately on per-VMT, the most impactful interventions are those reducing exposure, land-use changes, telehealth reducing medical-travel distances, or better local services, not necessarily road engineering. If a state scores poorly on both, as Mississippi does, road engineering, enforcement, and behavioral interventions are all indicated. Reading both metrics together, rather than relying on either alone, allows policymakers to target the right levers in the right places.