Skip to content
American Factbook
Menu

Method

How the motor-vehicle crash figures are made

How the pages of the Motor-vehicle crashes hub read what their publishers put out, and what that record can and cannot say. The method every topic shares — where a figure comes from, snapshots, freshness and corrections — is on the site’s methodology page.

What the crash file holds

One record for each fatal crash on a public road in the year, from NHTSA’s Fatality Analysis Reporting System. Each record carries the number of people who died within 30 days of the crash, the place it happened and the code of the county it is placed in. The file records no cause and assigns no blame.

What a death is here

A death is a person who died within 30 days of a crash on a public road, as the file counts it. Each state’s count is summed from NHTSA’s file, and the national count is the sum of the states’ counts; neither is NHTSA’s own published total, which this factbook has not yet checked them against. A fatal crash is one crash, and a crash can kill more than one person.

Counts withheld

The factbook withholds a county’s count of 1 to 10 deaths, and a county’s rural, urban or other-place split of 1 to 10, because a small count can identify the people who died. NHTSA suppresses nothing: the withholding is the factbook’s. Deaths are never fewer than crashes, so a printed crash count would bound or recover a withheld death count. A county whose deaths are withheld therefore has its crash counts withheld too, and a county with any death split withheld has all its death splits and crash splits withheld together; its total deaths and crashes still print where its total is not withheld. A withheld count is shown as withheld. It is never shown as zero, and it is not ranked.

State and national counts, and their splits, print in full, whatever their value, and every state is ranked. A count of 0 prints as 0, and so does a state’s count of crashes. A county with no placed crash has no row.

How the rankings are made

A state or county is ranked where its count of deaths is printed. Equal counts share a rank, and the next rank is skipped. The rank orders counts. It says nothing about risk.

Names and unmatched counties

A county is named from the Census Bureau’s county table, by its five-digit code (Census Bureau county population estimates). Connecticut keeps the county codes of its former counties in the file, and the Census table names its planning regions instead, so every Connecticut county is labeled not matched. Alaska county 02261 and South Dakota county 46113 are not matched either. A county not matched is labeled and its count is still printed and ranked. Nothing is dropped, and no crash is moved to another county.

No rates on these pages

The hub prints counts only. A rate per resident needs the population of each state and county, which is a later source, so no page divides a count by residents and no page compares places by their size.

What this data cannot show

  • A death is counted where a person died within 30 days of a crash on a public road. The file records no cause and assigns no blame.
  • The factbook withholds a county’s death count of 1 to 10, with its crash counts, and a county’s rural, urban or other-place split of 1 to 10, with all of that county’s splits, because a small count can identify the people who died and a crash count would bound it; NHTSA suppresses nothing. State and national counts are summed from NHTSA’s file and print uncut.
  • Connecticut, Alaska and South Dakota counties are not matched to the Census county names. They are labeled not matched, and their counts print and are ranked.
  • These are counts. A state or county with more driving has more crashes, and no population is used on these pages.

Sources

Data retrieved