The Fleet-Count Gap: How One Database Query Surfaces America's Ghost Fleets

The number nobody checks

When a motor carrier registers with the federal government, it writes down how many trucks it runs. That single figure — the power-unit count on the MCS-150 form — quietly drives almost everything downstream: the carrier's safety profile, its new-entrant audit, and the premium an insurer charges to put it on the road. It is self-reported, and it is almost never checked against anything.

We wanted to know what happens when that number stops describing reality. So we built a check and ran it across the entire country.

The check is one ratio:

For an honest one-truck operation, that ratio is about one. The same truck gets inspected, the same VIN repeats, and the number can't run away — a single truck can only be in one place at a time. When the ratio comes back 50, or 150, or 700, the reported number was never a count of equipment. It was measuring something else.

What we did

We computed this for 576,011 carriers that appear in the national roadside-inspection record (~5.8 million inspections over a rolling 24-month window), joined to the FMCSA company census (~4.4 million entities). Three things made the result trustworthy rather than noisy:

1. Normalization. VINs and plates were canonicalized (uppercased, punctuation stripped) so transcription variants aren't counted as different vehicles, and obvious placeholder VINs (all-zeros, "UNKNOWN," too-short strings) were discarded.

2. False-positive control. The legitimate reasons a small carrier might touch many VINs — rental and leasing fleets, manufacturer fleets, large for-hire carriers, government and transit agencies — were identified and set aside, 6,673 of them, each with a stated reason. They are real, they are not the story, and keeping them out is what makes the rest meaningful.

3. Screening, not verdicts. Every output is a signal that warrants review, not a finding. A reported fleet size is a point-in-time self-report; a genuinely growing carrier can show an elevated ratio innocently. That explains a ratio of five. It does not explain seven hundred.

What we found

After removing the legitimate high-VIN operators, the pattern is not a handful of outliers:

• 101 carriers reported one truck but were inspected on 50 or more distinct VINs.
• 43 of those crossed 100+ VINs on that single reported truck.
• 17 more reported 2–5 trucks while showing 100+ VINs.
• 342 carriers were inspected in 20 or more states while reporting just one or two trucks — a geographic footprint a one-truck operation cannot physically produce.
• The most extreme single carrier showed more than 700 distinct VINs against one reported truck.

These are not edge cases buried in the data. They sit at the very top of a ranked list of 1,000, and the carriers there share a profile: recently issued operating authorities, a national inspection footprint, and a reported fleet of one or two trucks.

It isn't a list — it's a network

The single-carrier ratios grab attention. The structural finding is what sits underneath them.

A VIN identifies one specific physical vehicle. So when the same VIN appears in inspections under two different carriers, the same truck operated under both authorities — a provable link between them. For one or two carriers, that can be innocent. But map every such link and the flagged carriers don't resolve into hundreds of unrelated one-truck firms. They resolve into connected webs, because the links chain together: carrier A shares a truck with B, B shares different trucks with C and D, and A, B, C, and D are now one structure even where A and D never directly touched.

Run that analysis nationally and 2,231 carriers connect into 155 distinct clusters — networks of small authorities through which the same equipment circulates. The connections are dense and concentrated: the carriers at the top of the list each share equipment with dozens to over a hundred other authorities, and within each cluster a few central nodes hold a disproportionate share of the shared vehicles — hubs the rest of the network draws from.

You can watch the network move. When one authority in a cluster is shut down, its trucks don't park — the same VINs reappear under other carriers in the same web, often within days. The authority on the paperwork changes; the iron keeps rolling.

This is not what a busy used-truck market looks like. A used-truck market scatters equipment randomly across the whole country and every fleet size. This is the opposite: the same vehicles circulating inside a tight, recurring set of small, recently-formed authorities, frequently clustered in the same metro areas and registered through the same handful of addresses. A formal test bears this out — the volume of equipment these carriers share comes in roughly 20 standard deviations above what random chance would produce. Three standard deviations is already treated as no coincidence.

Why would anyone structure operations this way?

We deliberately started from incentives rather than intent. Here is the first principle: in trucking, almost every adverse consequence is metered per operating authority. Insurance premium is priced off each authority's self-reported truck count and mileage. CSA safety scores and the interventions they trigger accumulate per authority. The 18-month new-entrant safety audit is per authority — let one lapse and get a fresh one, and the clock resets. Liability and the minimum insurance behind it attach to the authority that controlled the truck. Enforcement attention is allocated by each authority's record.

So a rational operator who wanted to run a large amount of equipment while minimizing every one of those metered consequences would do exactly what the data shows: spread the trucks across many small authorities, keep each one's reported size (and premium) minimal, keep each one's accumulated inspection record below the thresholds that trigger review, and abandon-and-replace any authority whose record degrades or draws enforcement. The behavior is the incentive-rational solution to "maximize equipment on the road, minimize per-authority exposure."

That model predicts everything we observe, which is why the behavior reads as strategic — not random and not merely operational. Four signatures point that way: equipment flows toward newer, cleaner authorities; moves cluster around enforcement events; the fleet-count underreporting is systematic across each network; and the same central nodes recur. The strongest specific incentives, in order, appear to be fragmentation of the safety record (so no single authority accumulates the history that triggers intervention), insurance premium arbitrage (a policy priced on one truck covering a fleet), and liability compartmentalization (thin, minimum-insured shells). None of this asserts criminal intent — that is a legal determination beyond inspection data. What the data settles is the question we set out to answer: the movement is structured, not incidental.

Why it matters beyond the regulator

It is tempting to file this under the agency's problem. It isn't, for three reasons.

Insurance is priced on the reported number. A premium built on one self-declared truck does not cover a fleet operating under that authority across dozens of states. When a loss lands, the carrier was priced as one truck and was running many — and the gap is ultimately absorbed across every honest fleet's premiums.

Brokers and shippers are newly exposed. A 2026 Supreme Court decision held that federal law does not preempt state-law negligent-selection claims against brokers. A broker who books a load with a carrier reporting one truck and running hundreds, on a safety record built under disposable authorities, can now be asked to answer for that choice. "We didn't check" stopped being a safe answer.

The honest operator pays for the rest — in premiums, in lane rates undercut by minimum-cost shells, and in the reputation of an entire segment.

The fix is a query

The most important finding is also the most encouraging one: the data needed to catch this already exists, in public records, and the check is cheap. Reconciling a carrier's reported truck count against the unique VINs in its inspection record is a database query that runs over millions of rows in minutes. We built one, validated it against an independent investigation that reached the same carriers, and extended it from a single documented network to 155 nationwide.

The point is not that a ratio is a verdict — it isn't. The point is that this reconciliation is almost never performed today: not at registration, not when an insurer binds a policy, not when a broker tenders a load. A ratio of ten vehicles to one reported truck is a reason to look closer. A ratio of a hundred and sixty to one is a reason to investigate before anyone signs.

The number on the form was never a count of trucks. For the overwhelming majority of carriers it is accurate and uninteresting. For a measurable few, it is the foundation of everything downstream — and it does not survive contact with the inspection record. Checking it is one query. The harder part, as always, is choosing to look.

---

This article reports aggregate findings from a screening analysis of public FMCSA inspection and census data. It names no carrier or individual and makes no determination about any specific company. Screening signals identify carriers whose inspection footprint is inconsistent with their reported fleet size and warrant review; they are not findings of wrongdoing.