Crash Rate Calculator — Intersections, Segments & Critical Rate

Rates, exposure and why the threshold moves

A crash rate divides crashes by exposure. For an intersection, exposure is the number of vehicles that enter it, in millions; for a segment, the vehicle-miles driven on it, in hundreds of millions. Three to five years of data is typical: shorter periods are noisy, and longer ones risk spanning changes in geometry or traffic.

Crash counts are random, roughly Poisson. A site with small exposure can show a high rate from one bad year. The critical-rate formula (the rate-quality control method) builds that in: the term K√(R_a/m) is the normal approximation to the Poisson spread, and 1/(2m) is a continuity correction. Both shrink as exposure grows, so a busy site needs only a small excess to be flagged while a quiet one needs a large excess.

The reference rate R_a should come from sites of the same type — for example, four-leg signalized urban intersections, or rural two-lane segments — usually published by the state DOT. Comparing a signalized intersection with an average for all intersections mixes populations and misleads.

Worked examples

Limits of this tool

• Crash rates assume crashes grow linearly with volume, which is rarely true. Safety performance functions and Empirical Bayes (HSM Part B) correct this and regression to the mean; use them where your agency has calibrated SPFs.
• Severity weighting (EPDO) and crash-type screening are not included.
• Use the same crash definition (reportable threshold, intersection-related radius) for the site and for the reference rate.

Frequently asked questions

How do you calculate an intersection crash rate?

R = N × 1,000,000 / (365 × Y × V), where N is the number of crashes in Y years and V is the average daily total of vehicles entering from all approaches. Twenty crashes in 3 years at an intersection with 24,000 entering vehicles per day is 20 / 26.28 = 0.76 crashes per million entering vehicles.

How do you calculate a segment crash rate?

R = N × 100,000,000 / (365 × Y × AADT × L), with L the segment length in miles. Thirty crashes in 3 years on a 2.4-mile segment carrying 12,000 vehicles per day gives 30 / 0.3154 = 95 crashes per 100 million vehicle-miles.

What is the critical crash rate?

The rate a site must exceed before it is considered significantly higher than the average for similar sites at a chosen confidence level: Rc = Ra + K√(Ra/m) + 1/(2m), where Ra is the average rate for the reference group, m the site's exposure (MEV or 100 MVMT) and K the normal deviate (1.645 for 95%). Sites with R > Rc are flagged for a detailed safety study.

Why not just rank sites by crash count?

Busy sites have more crashes simply because more vehicles use them. Rates normalize for exposure. Rates have their own bias — low-volume sites can show very high rates from one or two crashes — which is why the critical-rate test widens the threshold for small exposure. The Highway Safety Manual recommends Empirical Bayes methods with safety performance functions where they are available.

What volume should I use for an intersection?

The total entering volume: the sum of the daily volumes entering on every approach. Each approach carries roughly half of the two-way AADT on its leg, so where two two-way roads cross, the total entering volume is about the major-road AADT plus the minor-road AADT. Do not add exiting volumes.

References: AASHTO, Highway Safety Manual, 1st ed. (2010), Part B, ch. 4 (network screening: crash rate and critical rate methods). FHWA, Highway Safety Improvement Program Manual, FHWA-SA-09-029 (2010), §2.3. Garber, N.J. and Hoel, L.A., Traffic and Highway Engineering, 5th ed., Cengage (2015), ch. 5.