How We Calculate Crash Costs

These numbers are estimates, not invoices.

We calculate these costs using a method developed by the National Highway Traffic Safety Administration, or NHTSA. It uses one set of costs for crashes with injuries and another for crashes without injuries. The goal is to answer a simple question:

If these crashes happened here, how large was the human and financial toll?

Two Different Cost Views

On CrashCount district pages, we show two cost views.

• Direct money losses: things people, employers, insurers, and the public actually have to pay for or absorb, such as medical care, lost work, legal costs, travel delay, and property damage.
• Estimated human toll: the broader NHTSA estimate. This starts with the direct money losses and then adds the human cost of pain, disability, reduced quality of life, and lost years of life.

That is why the broader number is much larger.

Why The Numbers Can Look High

The biggest reason is simple: crashes with injuries cost far more than crashes with only vehicle damage.

A crash injury is not just an emergency-room bill. NHTSA says crash costs include “lost productivity, medical, legal and court costs, emergency service, insurance administration, congestion, property damage, and workplace losses.”

That means even a crash that does not look catastrophic at first glance can still create a large cost once you include:

• ambulance and emergency response
• hospital and follow-up medical care
• prescriptions and rehabilitation
• lost pay from missed work
• lost household work
• insurance administration
• legal and court costs
• employer and workplace disruption
• congestion and travel delay
• vehicle or property damage

The broader estimate goes even further. It adds the human loss that does not show up on a bill: pain, long-term disability, reduced quality of life, and lost life expectancy.

In everyday terms: a fender-bender is mostly about repairs and delay. A crash with injuries can also mean months away from work, physical therapy, lasting pain, and a life that never fully returns to normal.

The Source We Use

We use NHTSA DOT HS 813 403, Appendix D, which gives police-reported KABCO unit costs in 2019 dollars.

That matters because CrashCount is based on police-reported crash records. We are not using hospital billing data. We are using the federal cost schedule designed for police-reported injury severity.

How We Map NYPD Injury Reports

CrashCount stores NYPD injury severity in four public-facing harm buckets:

• Possible injury or pain complaint
• Visible injury, not reported as serious
• Suspected serious injury
• Death

These are the closest match to NHTSA’s police-reported KABCO groups:

• Possible injury -> C
• Visible injury, not reported as serious -> B
• Suspected serious injury -> A
• Death -> K

This is a police-report severity system, not a hospital diagnosis.

The Unit Costs We Use

For this first release, we keep the NHTSA values in constant 2019 dollars.

How The Math Works

For each district and time window:

1. We count the number of people in each injury bucket.
2. We multiply each count by the matching NHTSA unit cost.
3. We add those subtotals together.
4. For crashes where nobody was reported hurt, we count the damaged vehicles in those crashes and apply NHTSA’s damage-only unit cost per vehicle.

In plain language:

district total = injured people by severity × federal unit cost, plus damaged vehicles in no-injury crashes × federal damage-only cost

How To Read The Card

The card breaks the estimate into three parts:

• People hurt or killed: this is the sum of the first four rows in the table.
• Damage-only crashes: this is the damage-only vehicle row.
• Overall total: this is people hurt or killed + damage-only crashes.

The card also shows two valuation styles for each part:

• Direct money losses
• Broader estimate

These are two different ways to value the same crashes. They should not be added to each other.

Another way to read the same math is:

direct money losses + added harm in the broader estimate = total estimate

On the card, that middle “added harm” amount is simply:

broader estimate - direct money losses

Why We Count Damage-Only Vehicles, Not Just Crashes

NHTSA’s damage-only unit in Appendix D is a per-vehicle number, not a per-crash number.

So if a no-injury crash involved two damaged vehicles, that counts as two damage-only units in our math.

CrashCount now uses NYPD’s vehicle-damage fields to count those vehicles. We do not use all vehicles in a no-injury crash by default. A vehicle has to show reported damage in the NYPD record to count toward this damage-only part of the estimate.

This is the most source-backed way to use the federal schedule.

What You Can Download

Every district card includes downloadable CSV and JSON files showing:

• the counts used
• the unit costs used
• the multiplication for each row
• the subtotal for each row
• the run ID and method version used

That means every public total can be checked line by line.

Important Limits

• These are estimates, not actual payments by a district, agency, or insurer.
• The broader estimate is intentionally larger because it includes pain and lost quality of life.
• Different map layers should not be added together. The same crash can appear inside more than one district system.
• We currently keep the federal schedule in constant 2019 dollars. We are not yet inflating it year by year.

Read The Federal Source

• NHTSA DOT HS 813 403

The key sections for this project are:

• Appendix D for police-reported KABCO unit costs
• the report’s explanation that the broader estimate adds quality-of-life loss to direct economic loss