Research Study

Key Findings

Concurrent measures from the February 2026 snapshot: violations and crashes in the same 24-month window, exposure-adjusted but not controlled for inspection activity. The crash-probability figure is a share of motor carriers, not adjusted for fleet size.

Predictive Validation (Year 1 → Year 2)

From the study’s own rate-based tables below (2024 predictors, 2025 crashes); relative risks compare motor carriers with the predictor against motor carriers without it.

Methodology

We analyzed FMCSA Safety Measurement System (SMS) violation data, linking specific violation types (Group_Desc) to motor carrier crash records. Crash rates were calculated using Empirical Bayes adjustment to account for exposure differences between motor carriers.

Data Sources

• Violations: SMS_Input_Violation file (6,548,807 records)
• Crashes: SMS_Input_Crash file
• Motor Carrier Census: FMCSA Motor Carrier Census
• Inspections: Roadside inspection records

Statistical Approach

For each violation type, we calculated:

• Relative Risk: Crash rate for motor carriers with the violation type divided by crash rate for motor carriers without, both measured in the same 24-month window (a concurrent measure, not controlled for inspection activity)
• Crash Probability: Percentage of motor carriers in each group that experienced at least one crash (a share of motor carriers, not adjusted for fleet size)
• Dose-Response: Crash metrics stratified by violation count (0, 1, 2, 3-4, 5-9, 10+)

Relative Risk by Violation Type

The following table shows crash rate relative risk for motor carriers with each violation type compared to motor carriers without that violation type.

Dose-Response Analysis

A critical finding is the clear dose-response relationship: as the number of violations increases, crash probability increases monotonically.

Speeding Violations

Key insight: Motor carriers with 10+ speeding violations averaged 132x the fatalities per motor carrier (0.5963 vs 0.0045) of motor carriers with zero speeding violations. That is a concurrent, per-motor-carrier count rather than a rate per unit of exposure, and fleet size drives most of it: the exposure-adjusted crash rate in the same table rises 1.2x (0.0477 to 0.0570). The score has no fatality-by-violation term; a fatal crash counts in the motor carrier’s crash record like any other crash, and its being fatal enters only as a screen (it withholds Excellent and is one of the events that confirms a Critical grade).

Dangerous Driving Violations

Behavioral vs Equipment Violations

A key finding is that driver behavioral violations are significantly stronger predictors of crash likelihood than equipment/vehicle violations.

Statistical Significance

All findings are statistically significant at p < 0.001 given the large sample sizes. Correlation coefficients between the study’s own violation metrics and concurrent crash rates (none of these is used as measured here; the score reads HOS violations inside the driver-behaviour relativity, out-of-service inspections and violations of every kind through the standardised out-of-service and all-violations ratios, and inspection activity through the class rate and the proof screens, as described below):

How FRED Uses These Findings

The FRED Score reads a motor carrier’s violations through three relativities, each the motor carrier’s own ratio shrunk, with a credibility weight, toward what fleets of its size and inspection mix usually show, and multiplies them, each raised to a fitted exponent (0.112 for driver behaviour, 0.083 for the out-of-service ratio, 0.077 for all violations), into the crash rate the class model expects before the motor carrier’s own crash record is weighed:

• Driver behaviour — Unsafe Driving and Hours-of-Service violations per driver-level inspection over 24 months, against the rate for fleets of the motor carrier’s size; credibility n / (n + 10) inspections. Under FMCSA’s BASIC assignment, speeding, reckless and dangerous driving, seat-belt, handheld-phone and texting violations are Unsafe Driving violations, and false logs and driving fatigued are Hours-of-Service violations, so this is where the behavioral types studied here count.
• Out-of-service ratio — out-of-service inspections against the number expected for the states and inspection levels under which the motor carrier’s inspections were reported (an inspection by a federal inspector is filed under a federal code rather than a state, so it is compared with federal inspections of the same level wherever they took place); credibility e / (e + 1) expected events.
• All violations — violations of every kind against the number expected for the same inspection mix; credibility e / (e + 3) expected events.

No violation type has a term of its own: not speeding, not reckless driving and not drug or alcohol violations. The Controlled Substances/Alcohol BASIC was tested as a rate term of its own and rejected as too rare to pass the study’s stability test across inspection facilities, so a drug or alcohol violation counts only through the all-violations ratio and, when it puts the driver out of service, the out-of-service ratio (which counts inspections with any out-of-service violation), never as a term of its own; and a count of speeding violations (10+, 5–9) is not an input. The grade is the band of the motor carrier’s relativity to its class rate, so a worse violation record moves the grade through those three terms in proportion to the evidence behind it, not through a fixed deduction per violation type. The exponents set the leverage: a fully credible driver-behaviour ratio of twice the class rate raises the expected crash rate by about 8 percent rather than doubling it.

The score does apply grade ceilings, and one of them is driven by exactly the behavioral violations studied here: Excellent is withheld when the driver-behaviour relativity is above 1.0 (worse than class), alongside its other proof screens (a posterior probability of at least 0.6 that the motor carrier’s relativity sits below the Excellent cut of 0.60, at least three inspections, at least 0.1 inspections per filed power unit, no fatal crash in the window and at least 0.3 expected crashes). A single truck is capped at Strong. A current FMCSA Unsatisfactory or Conditional rating caps the grade at Marginal, with the score at most 50 and a referral. An interim exposure cap moves an Excellent to Strong and a Strong to Satisfactory when a fleet of 10 or more filed power units has had fewer than a quarter of them seen roadside in 24 months and fewer than 0.25 inspections per filed unit-year. And an improvement in grade not caused by decay roll-off, a data correction or a recalibration is ramped one band per run; a worsening is never held back.

Crash Probability by Behavioral Flag (February 2026 snapshot)

The study’s own concurrent measure: the share of motor carriers carrying each flag that had at least one crash in the same 24-month window. The shares are not adjusted for fleet size or inspection activity, and the flags themselves are not inputs to the score:

The temporal validation below is the study’s evidence that behavioral violations predict future crashes, not only concurrent ones. Measured forward and controlled for inspection activity, violation detail adds little to crash-frequency prediction once inspection and out-of-service counts are known; its job in the score is the driver-behaviour relativity and the proof screen for Excellent.

Predictive Validation: Year 1 → Year 2

To validate the predictive power of violation types, we conducted a temporal split analysis: using 2024 data to predict 2025 crashes. This approach tests whether behavioral violations truly predict future crashes, not just concurrent ones.

Key Question: Do Behavioral Violations Predict Future Crashes Better Than Past Crashes?

A concern with using past crashes as a predictor is size bias—larger motor carriers have more opportunity for crashes simply due to exposure. Behavioral violations may provide a more size-independent signal of underlying risk.

Relative Risk by Violation Type (Rate-Based, Size-Controlled)

By using crash rates (per 100k miles) instead of counts, we control for motor carrier size:

Size-Controlled Analysis: Violations Within Fleet Size Bands

When we analyze within fleet size bands, behavioral violations become dramatically more predictive. This reveals predictive power that's hidden when mixing motor carriers of different sizes:

Cumulative Effect: Multiple Behavioral Violation Types

Motor carriers with multiple types of behavioral violations show dramatically compounding risk:

The score itself is not validated by this Year 1 → Year 2 classification: it is checked on every run against out-of-time crash outcomes by fleet-size cohort (exposure-weighted Gini and calibration gates), as described on the Empirical Validation page.

Conclusion

This study demonstrates that specific violation types, particularly behavioral violations related to driver decisions, are significantly stronger predictors of crash likelihood than aggregate violation counts or equipment violations alone.

The dose-response relationship—where more violations of high-risk types correspond to dramatically higher crash and fatality rates—provides empirical support for treating driver-behavior violations as a risk signal in their own right and not only as part of a single violation count, which is how the FRED Score treats them.

Key Findings

From the concurrent analysis of the February 2026 snapshot (shares of motor carriers with at least one crash in the study window, not adjusted for fleet size):

• Motor carriers with reckless driving violations have a 71% crash probability
• Motor carriers with 10+ speeding violations have a 98% crash probability
• Motor carriers with 10+ speeding violations averaged 132x the fatalities per motor carrier of those with none — a per-motor-carrier count driven largely by fleet size; the exposure-adjusted crash rate rises 1.2x
• In the FRED Score these violations enter one driver-behaviour relativity (Unsafe Driving and Hours-of-Service violations per driver-level inspection, shrunk toward the violation rate of fleets its size with a credibility of n / (n + 10) inspections) alongside the standardised out-of-service and all-violations ratios; no violation type has a term of its own, and a motor carrier whose driver-behaviour relativity is above 1.0 cannot be graded Excellent

Predictive Validation: Year 1 → Year 2

The Year 1 → Year 2 temporal split shows that the violation signals carry forward:

• A 2024 crash remained the strongest single predictor of a 2025 crash (1.67x relative risk); reckless/dangerous driving (1.31x), distracted driving (1.27x) and speeding (1.23x) each predicted 2025 crashes on their own
• Behavioral violations predict crashes independently of motor carrier size:
  • Small motor carriers with substance violations: 2.75x relative risk
  • XLarge motor carriers with speeding/reckless: 3x relative risk
• Motor carriers with 3+ behavioral violation types have 65% crash probability the following year

The score keeps violation detail at the level this study supports: driver-behavior violations form a relativity of their own, and Excellent is withheld when that relativity is worse than class. Measured forward and controlled for inspection activity, violation detail adds little to crash-frequency prediction once inspection and out-of-service counts are known, so its job in the score is that behaviour term and the proof screens, not a per-type weight. The temporal validation confirms that these signals are truly predictive—not just correlated with concurrent outcomes.