Cross-Border Natural Experiment
The Inspection Paradox
Why inspection-based safety scores penalize the most-scrutinized carriers — not the most dangerous ones
Published: May 2026 • 1.07M carriers with estimated ISS scores
Key Findings
The FMCSA's Inspection Selection System (ISS) — and inspection-based safety scores like it — measure the violations a carrier accumulates at roadside. But violations are only recorded when a truck is inspected. We use a natural experiment — carriers domiciled in Mexico, Canada, and the US, who face radically different inspection intensities — to test whether ISS measures crash risk or simply inspection exposure.
The paradox: ISS flags Mexican carriers for inspection more than three times as often as US carriers and more than twice as often as Canadians. Yet Mexican carriers crash less than either group. The carriers ISS treats as the highest priority for inspection crash at roughly the same rate as the carriers it clears to pass. ISS is reading inspection intensity, not road risk.
The Natural Experiment
Carriers crossing the southern border are funneled through fixed ports of entry, where vehicle inspections are frequent and systematic. Canadian carriers face far lighter border scrutiny, and US domestic carriers are inspected only occasionally at random roadside locations. That gives us three populations doing the same job — hauling freight on US highways — under three very different inspection regimes.
| Per-carrier average | Canadian | US Domestic | Mexican |
|---|---|---|---|
| Roadside inspections | 12.5 | 3.9 | 56.3 |
| Vehicle Maintenance measure | 2.67 | 2.14 | 9.19 |
| Unsafe Driving measure | 0.82 | 0.91 | 0.37 |
| Hours-of-Service measure | 0.42 | 0.26 | 0.11 |
BASIC measures are time- and severity-weighted violation rates per relevant inspection (FMCSA SMS methodology).
Mexican carriers' Vehicle Maintenance measure is more than triple everyone else's — equipment violations are exactly what intensive border inspection surfaces. Yet on the driver-behavior BASICs that border inspections don't manufacture (Unsafe Driving, Hours-of-Service), Mexican carriers look cleaner than both US and Canadian carriers.
How ISS Sorts the Three Groups
ISS assigns every carrier to Inspect (high priority), Optional (inspector discretion), or Pass (low priority), driven largely by the Vehicle Maintenance and other roadside BASICs. With Mexican carriers carrying triple the equipment-violation load, ISS flags them overwhelmingly:
ISS Recommendation by Domicile
Share of scored carriers in each ISS bucket
Average ISS score: 68.4 Mexican, 45.5 Canadian, 37.5 US. On paper, ISS rates the Mexican fleet as nearly twice as risky as the US fleet.
Does “Inspect” Actually Mean “Crashes More”?
It's a fair question — and within a single inspection regime, ISS holds up. Pooling all carriers, those ISS flags “Inspect” crash about 2.8× as often per mile as those it clears to “Pass.” ISS is not noise.
| All carriers, by ISS bucket | Carriers | Crashes / 100k mi |
|---|---|---|
| Inspect | 214,272 | 0.0217 |
| Optional | 41,608 | 0.0220 |
| Pass | 810,065 | 0.0078 |
So ISS works when everyone is inspected at the same intensity. The problem appears the moment you compare populations that are not.
…But the Label Doesn't Travel Across Borders
Break the same numbers out by domicile and the ISS label loses its meaning. A Mexican carrier flagged “Inspect” crashes at essentially the same per-mile rate as a Canadian carrier ISS clears to “Pass,” and below a US “Pass” carrier.
| Domicile & ISS bucket | Carriers | Crashes / 100k mi | Crashes / 100 trucks |
|---|---|---|---|
| 🇲🇽 Mexican — Inspect | 3,509 | 0.0059 | 0.51 |
| 🇨🇦 Canadian — Pass | 6,193 | 0.0060 | 0.22 |
| 🇺🇸 US — Pass | 802,263 | 0.0078 | 0.065 |
| 🇺🇸 US — Inspect (for reference) | 207,746 | 0.0221 | 2.19 |
Within every domicile, “Inspect” carriers do crash more than “Pass” carriers — but the levels are set by inspection regime, not risk. A US “Inspect” carrier (0.0221) sits far above a Mexican “Inspect” carrier (0.0059).
Aggregated across all buckets, the cross-border verdict is unambiguous and holds on both exposure bases: Mexican carriers are flagged “Inspect” 2.3× more often than Canadians, yet crash ~45% less — 0.0057 vs. 0.0101 per 100k miles, and 0.40 vs. 0.77 per 100 trucks. ISS ranks them in exactly the wrong order.
Why This Happens
ISS is built on the SMS BASIC percentiles, which are dominated by violations discovered during inspection. A violation can only be recorded if a truck is stopped and examined. So the metric has a built-in dependence on how often a carrier is inspected:
- Inspect a fleet 14× more often and you will find 14× more equipment defects — even if its trucks are no worse maintained.
- Equipment defects found at the border (a worn tire, a marker lamp out) are real, but most never cause a crash. They inflate the Vehicle Maintenance BASIC without moving the crash rate.
- Lightly-inspected carriers accumulate few recorded violations and drift toward “Pass” — regardless of how they actually drive.
The result is a score that conflates scrutiny with safety. It is internally consistent within one inspection environment but cannot be compared across carriers inspected at different intensities — which is precisely what insurers, brokers, and shippers do every day when they line a Mexican carrier up against a domestic one.
What a Crash-Predictive Score Has to Do Instead
If the goal is to predict crashes — the outcome that actually drives loss — an inspection-violation tally is the wrong instrument. A defensible safety score has to:
Measure events against miles driven and inspection opportunity, so a heavily-inspected carrier isn't penalized for the simple fact of being looked at more often.
Weight each signal by how much it actually predicts a future crash — not by how many times it gets written up — and shrink thin records toward the population norm rather than over-reacting to them.
That is the principle behind Fleetidy's scoring methodology: a forward-looking, exposure-normalized, crash-calibrated model rather than a roadside-violation count.
Methodology & Caveats
ISS is estimated, not official. FMCSA does not publish carrier-level ISS values. We reconstruct them with the published FMCSA ISS-CSA Safety Algorithm (Dec 2012), deriving BASIC percentiles from the raw SMS measures FMCSA does release. Validated against the carriers where official ISS data is available, the reconstruction agrees on the Inspect / Optional / Pass bucket 95.6% of the time.
Two exposure bases, one conclusion. About 69% of Mexican carriers have mileage we judge unreliable, so their per-mile rates rest partly on imputed exposure. We therefore report crashes per 100 trucks alongside crashes per mile. The headline cross-border comparison (Mexican carriers crash less than Canadian carriers despite far heavier ISS flagging) holds on both bases.
ISS is not worthless — it is not portable. Within a single inspection regime, ISS ranks crash risk in the right direction. The failure is cross-regime: it cannot fairly compare carriers inspected at 56 vs. 4 inspections each. Our critique is of that misuse, not of within-regime triage, which is what ISS was designed for.
Observational data. Differences reflect FMCSA records as reported. Crash and mileage reporting completeness may vary for carriers operating primarily outside the US, and small sub-groups (e.g., Mexican “Optional,” n=21) are shown only for completeness.