Reading the evidence · 2026-10-27 · BestCityToLive Research

How to read a metropolitan road-death rate

A regional fatality count can inform a city comparison. Its boundary, crash definition, population denominator and release status determine what it can tell a traveller or resident.

Start with the source region

A metropolitan statistic can cover a central city, suburbs and outlying counties. U.S. metropolitan statistical areas use counties or county equivalents as building blocks; their membership reflects the official delineation and its reference period. A familiar city name does not make a metropolitan figure a measurement of its downtown, a neighbourhood or a particular walking route.

Keep the original region name and identifier with the number. Use a compatible area for the crash count and population denominator. The same area name across two releases does not prove that its constituent counties stayed the same.

Keep the crash definition attached

NHTSA’s Fatality Analysis Reporting System records fatalities in its defined motor-vehicle traffic crashes. Its case definition requires a motor vehicle in transport and uses a 30-day follow-up convention. Victims can include pedestrians and cyclists as well as vehicle occupants; that does not turn the file into a census of non-motor-only crashes.

The trafficway convention concerns the source’s public-access traffic scope. It should not be replaced with a simple test of who legally owns the road. A different transport or health dataset may use a different event definition or follow-up period. Those differences need review before figures are combined.

Separate people, crashes and calendar years

One crash can kill more than one person. A fatality rate therefore needs the count of qualifying people, rather than a count of fatal crashes. Preserve the source classification and its unknown codes; an unreported injury category is not an observed zero or a confirmed death.

A crash-year series can include a qualifying death after the calendar year ends. Selecting only people whose recorded death year matches the crash year can remove valid cases. Where individual dates are unavailable, retain that limitation instead of inventing a timestamp or claiming that every follow-up interval was independently reconstructed.

Use a denominator that answers the same question

Dividing territorial fatalities by midyear residents gives a population-normalised regional count. Visitors and people travelling through the region can be part of the numerator. Residents are not the same denominator as journeys, distance travelled or time spent on the road.

The result cannot establish an individual traveller’s likelihood of injury, compare walking with driving, or identify a safer neighbourhood. Those questions require compatible mode, location and exposure evidence. Population estimates can also be revised: retain their reference date and release vintage alongside the rate.

Treat a revision as evidence, not a new observation year

An Annual Report File and a finalised release can differ. A newer download date does not make the crashes more recent. Record provisional status, the original period and the exact source version. A comparison using a later population vintage must keep its changed denominator and geographic assumptions visible.

A regional source desk is useful when a reader can follow the count, denominator, boundary and qualifications back to the originals. Cite those details together. A single historical measure does not establish an overall city score or a current visitor-safety ranking.

Sources & scope

Citation: BestCityToLive Intelligence / AINT.app (2026-10-27). How to read a metropolitan road-death rate. bestcitytolive.com/research/reading-metropolitan-road-death-rates.

Editorial & research standards →

Research interpretation by BestCityToLive. Source observations may refer to earlier periods. This note does not constitute an official Civometra Index results release.

Explore the index Compare cities