Data Sources & Methodology

Core Definitions

Defined terms (§1–§7)

Cohort Definitions

Every SMS BASIC percentile rank partitions the carrier population by 5 × 4 = 20 cohort buckets: fleet-size × safety-event-count.

Both axes define the cohort partition for SMS BASIC percentile rank (since FMCSA's public files publish raw MEASURE values only — see Limitations). Carriers in the 0-events bucket receive cohort-baseline percentiles regardless of their MEASURE because there's no observable behavior to rank.

Data Sources

All inputs are public FMCSA records. Snapshot timestamps below auto-populate from the ingestion audit log; each refresh updates the corresponding row.

Bulk FMCSA Source SMS Data files refresh monthly. The L&I bulk file (ActPendInsur) refreshes daily on catalog.data.gov. We re-ingest on the documented cadence; new monthly snapshots are loaded within ~7 days of FMCSA publication.

Night-Driving Pattern Detection

Per-carrier nocturnal-clustering identification is performed under the SafeNY Night-Driving Pattern Detection Framework, an in-house implementation that combines a one-sample binomial significance test against the national dark-condition baseline with ARCH-family conditional-variance heuristics for temporal-cluster identification (Engle, 1982 — see References §4). Each carrier's ~2-year crash record is classified by Light_Condition_Desc and tested against the national dark-condition crash share. The page surfaces the share and a one-sample binomial p-value when the carrier has ≥5 documented crashes in the window; below 5, the section reports insufficient sample.

Attribution. This night-driving pattern-detection framework is a SafeNY-developed module. It does not modify the cited primary econometric estimators — it composes them. The framework name is internal nomenclature, not a citation to a single external paper; the underlying estimators (binomial test via normal approximation; ARCH conditional-variance heuristic) are standard and drawn from peer-reviewed econometric literature listed in the References section.

Data source

FMCSA SMS Crash File 2026-04 Snapshot, field Light_Condition_Desc. Approximately 259k national records; 99.6% non-null on the field. Source values fall into eight categories: Daylight, Dawn, Dusk, Dark - Lighted, Dark - Not Lighted, Dark - Unknown Roadway Lighting, Unknown, Other.

Night definition

Night aggregates the three Dark-* categories: Dark - Not Lighted, Dark - Lighted, Dark - Unknown Roadway Lighting. National combined share: 25.7% (the binomial baseline used in the per-carrier test).

Excluded from night: Dawn (2.4% national) and Dusk (1.4%). These transition periods carry separately elevated risk in the NHTSA Large Truck Crash Causation Study literature, but the underlying mechanisms (glare, reduced contrast during a brightness transition) differ from sustained-darkness fatigue and are not collapsed into the same metric.

Baseline

The 25.7% baseline is computed directly from the source distribution in the same dataset we use for per-carrier measurement. The dataset is both the measurement source and the comparison reference — internal consistency by construction. No external estimate is imported.

Significance test

One-sample binomial against the 25.7% baseline, two-sided. The p-value is computed via normal approximation with Postgres's erf() (Abramowitz & Stegun erf-form of the standard normal CDF). Sufficient fidelity for decision-support; not a publication-grade significance test, particularly for the smallest qualifying samples.

Sample-size gate

Carriers with fewer than 5 documented crashes in the ~2-year window get an insufficient sample message instead of a share + p-value pair. The 5-crash floor is conservative — at smaller sample sizes the normal approximation degrades and an apparent elevation is more likely noise than signal.

Regulatory connection

49 CFR § 392.14 requires extreme caution when hazardous conditions (including reduced visibility) adversely affect visibility or traction. The federal Hours-of- Service regime at 49 CFR § 395 limits consecutive driving hours and mandates rest periods; sustained-darkness over-representation in a carrier's crash record can correlate with fatigue-window operations.

Limitations

Long-haul carriers may have legitimately higher dark-driving exposure because their operations type concentrates over overnight hauls. The exposure-type confound is not corrected for here. The page displays deviation and significance together precisely so the reader can judge whether elevation reflects an operational pattern or a safety pattern — this is an investigative signal, not a fault determination.

Self-Insurance Detection (BMC-91X)

FMCSA permits large motor carriers to self-insure their bodily injury & property damage (BIPD) exposure rather than carry a third-party policy. The regulatory hook is 49 CFR § 387.309 (Self-insurance authorization for motor carriers and freight forwarders). Self-insurers file via the BMC-91X form and must demonstrate financial capacity to FMCSA on an ongoing basis.

Why the public dataset looks like "$0 coverage"

In the FMCSA L&I bulk feed, self-insured BIPD filings appear with a third-party liability_limit of $0. The carrier is not without coverage — they have regulatory authorization to satisfy claims from their own balance sheet — but the bond amount underwriting that authorization is not exposed in the public dataset.

Detection rule

A carrier is presumed self-insured when all three conditions hold against the most recent ingest:

1. Fleet size power_units ≥ 100 — floor that distinguishes BMC-91X self-insurers from the residual multi-subsidiary filing pattern.
2. An active BIPD filing on record (coverage_type starts with BIPD, is_active = true).
3. The filing's liability_limit is $0 or NULL.

The rule is implemented as a Postgres function public.apply_self_insurance_detection() invoked after each monthly recompute so the classification is refreshed on each ingest cycle.

What changes on the carrier page

• confidence_level = 4 (the regulatory authorization is itself a high-confidence signal).
• underinsured_flag = FALSE — self-insurance is not under-coverage.
• Coverage Gap is not computed; the public dataset does not expose the bond amount underwriting the authorization.
• Insurance section label reads Self-insured (BMC-91X, presumed).

Sample size + limitations

Approximately 92 carriers match the detection rule on the current snapshot, including FedEx Corp, Walmart Transportation, Old Dominion, Schneider National, Estes Express, XPO Logistics, Penske Logistics, and the FedEx Freight Inc operating subsidiary. The count refreshes on each ingest cycle as filings change.

The (presumed) qualifier is intentional: we infer the regulatory status from the data pattern (zero-limit BIPD on a large fleet), not from a direct FMCSA self-insurance roster call. A small residual set of multi-subsidiary carriers (parent corporation files BIPD under a different USDOT) may still produce the zero-limit pattern at fleet sizes below the 100-PU floor; those carriers stay outside the self-insured set and continue to surface the multi-subsidiary narrative from v6.4.1.

Evidence Density Score (EDS)

The Evidence Density Score reflects the corroboration depth of a carrier's federal regulatory record — the extent to which its safety posture is documented across multiple, mutually-reinforcing primary FMCSA sources. It draws on three record dimensions: inspection history, crash-record consistency, and authority continuity. EDS is a measure of how thoroughly the federal record corroborates a carrier's safety profile; it is not a prediction of liability, fault, settlement value, or any specific case outcome.

Record dimensions

• Inspection history — the volume, recency, and BASIC-domain coverage of roadside inspection and violation records, which establish how consistently the carrier's operations are documented in the federal record.
• Crash-record consistency — the recency-weighted presence of fatal- and severe-injury crash records in the FMCSA Crash file, with recent records weighted more heavily than historical ones (records remain within the discoverable retention horizon for roughly a year).
• Authority continuity — operating-authority and insurance-filing continuity, which indicates how stable and traceable the carrier's regulatory footprint is over time.

Categorical bands

EDS is reported as one of three categorical bands. The band, not a numeric score, is what is surfaced on carrier pages:

Stability under perturbation

Sensitivity analysis: the categorical band assignment is robust to boundary perturbation — top-cohort band membership retains 96–100% overlap when the underlying component weights are perturbed by ±10–25%. EDS is therefore not load-bearing on any single weight — a Daubert-style robustness property.

Stated limitations

• The component weights are heuristic priors, not empirical maximum-likelihood estimates; each is published so an independent expert can reconstruct the band assignment under Fed. R. Evid. 702 testability standards.
• The jurisdiction-tier component assumes within-state evidentiary-record homogeneity; sub-state venue variance (e.g., Los Angeles County versus rural Northern California, both Tier A) is not modeled.
• Record corroboration depth is independent of insurance posture — a self-insured carrier (BMC-91X) with a dense, recent federal record can still fall in the HIGH band, since discoverable-record density does not depend on the bond instrument.
• The crash-record recency weighting assumes uniform decay across crash severities. A severity-conditional weighting (multi-fatality incidents decaying more slowly) is left for Phase 2 calibration.

Reproducibility

An independent reviewer using only public FMCSA data can recompute every published figure within rounding error. The pipeline is open in structure; the only non-public surfaces are operational (database hosting, refresh-cadence orchestration). All formulas in the sections above are implemented directly in the deterministic stages below. The section documents data sources, computational environment, and fixed numerical parameters in the form required for Fed. R. Evid. 702 / Daubert testability.

Primary data sources

• FMCSA SAFER — Motor Carrier Census, monthly snapshot. Source: safer.fmcsa.dot.gov (public).
• FMCSA SMS — Safety Measurement System monthly files (Crash, Inspection, Violation, SMS_AB_PassProperty, SMS_C_PassProperty). Source: ai.fmcsa.dot.gov/SMS/Tools/Downloads.aspx (public).
• FMCSA L&I — Licensing & Insurance ActPendInsur, daily delta. Filename: actpendins_allwithhistory.txt. Source: catalog.data.gov (public).

Computational environment

• Language. Python 3.11. ETL and statistical computation entirely in standard scientific-stack libraries: pandas, numpy, scipy.stats, statsmodels (binomial test, ARCH conditional-variance fitting, LogNormal MLE).
• Storage. PostgreSQL 15 (Supabase). All calibration SQL is checked into supabase/migrations/ — 28 migrations as of the current methodology release.
• Orchestration. Staged execution script pipeline/scripts/recompute_staged.py runs the full pipeline as 13 deterministic stages; each stage is independently re-runnable and produces identical output for identical input.
• Determinism. No random-seeded estimators are used. All shrinkage and significance-test estimators are deterministic given input data plus the fixed parameters below.

Fixed numerical parameters

Recompute procedure (deterministic stages)

1. Download FMCSA SMS source for the target month from ai.fmcsa.dot.gov/SMS/Tools/Downloads.aspx. Six files: Motor Carrier Census, Crash, Inspection, Violation, SMS_AB_PassProperty, SMS_C_PassProperty.
2. Download L&I active filings actpendins_allwithhistory.txt from catalog.data.gov.
3. Compute SMS percentile ranks per cohort. Public SMS files publish raw MEASURE values; apply PERCENT_RANK partitioned by (fleet_cohort, safety_event_count_bucket).
4. Compose EDS: evaluateF(c, t) with the 180-day half-life kernel, then multiply through the seven M_* factors. Apply theM_feedback ≤ 5.0 clamp.

Reference SQL for the published computations is in our public migration files (supabase/migrations/). The literature underlying each numerical choice above is cited in the References section.

References

The estimators and statistical procedures used throughout this methodology are drawn from the following primary literature. Cited in APA 7 format.

1. Engle, R. F. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica, 50(4), 987–1007.Original ARCH paper. Provides the conditional-variance framework relied upon by the SafeNY night-driving pattern detection module for nocturnal-clustering identification (see §Night-Driving Pattern Detection).
2. Abramowitz, M., & Stegun, I. A. (Eds.). (1972). Handbook of mathematical functions with formulas, graphs, and mathematical tables (10th printing). National Bureau of Standards.§7.1.26 erf rational approximation; used inside the normal-CDF computation underlying the binomial significance test in the night-driving pattern section.
3. American Transportation Research Institute (ATRI). (2023). Understanding the impact of nuclear verdicts on the trucking industry. ATRI Research Reports.Source for state-level verdict-severity tier ratios used to calibrate M_state in EDS.

Disclaimer for Ranking Data

Disclaimers

• Not legal advice. The compiled FMCSA figures on this site are decision-support information for case selection. They are not, and do not substitute for, legal advice. Consult a licensed attorney for case-specific evaluation.
• Statistical record only. These figures describe a carrier's federal safety record from public records; individual case outcomes vary widely with facts, jurisdiction, evidence admissibility, and counsel quality. We do not predict any specific case's verdict.