Methodology | Freight Capacity Exit Index (FCEI)

What it measures

A carrier that stops running stops paying for insurance, and the insurer files that cancellation with FMCSA. We count the carriers that had such a filing, once each however many filings they had, against the insured carrier base domiciled in the same state. Both sides of the rate are counts of carriers, so nothing is compared against a different unit.

That is a different measurement from a failure count. A failure tracker counts carriers whose collapse became a public event, which is a small and late subset of the ones that leave. Most carriers are one or two trucks and they go quietly. A cancellation gets filed for every insured carrier whatever its size, and it gets filed when the coverage ends rather than months later when the paperwork catches up. FMCSA publishes the filings. It does not publish an exit rate built from them.

It is a consistent sample and not a census. Rates compare across months and states. They are not national totals.

Where the data comes from

Published federal records, and nothing else. No customer of FreightScout appears in any figure on this site. Four FMCSA files do the work, and each one supplies a different thing. They are named separately here because a reader checking a number needs to land on the file that holds it.

InsHist, the insurance filing history. Every policy that has taken effect and every policy that has been cancelled, with the dates. The cancellations are the numerator of the exit rate, and the effective dates are half of the coverage panel underneath it. Refetched monthly.

ActPendInsur, the active and pending filings. Policies currently in force or awaiting processing. This is the other half of the coverage panel, and it is what stops a carrier being counted as gone while its replacement filing is still in the queue. Refetched monthly.

Rejected, the refused filings. A filing an insurer submitted that FMCSA would not accept. This is the rejection rate, and it is the reason the capture schedule matters: FMCSA serves roughly a year of it and drops the trailing edge for good, so a month nobody captured cannot be bought back later. We hold an archive of every capture. As of September 2026 this file has published nothing newer than 14 May 2026, which is stated here rather than left for a reader to discover.

Carrier, the authority file. Used for one thing only: which state a carrier is based in, so a filing can be counted against the right state. It carries no insurance data and none is taken from it. It is also replaced in place with no date column, and it stopped changing for a fortnight in August 2026, so where it is silent about a carrier the domicile is read from FMCSA's newer Motus carrier file instead.

Facility counts, where a page shows one, come from the EPA Facility Registry Service, which lists environmentally permitted sites rather than shippers.

Cite as

FreightScout, Freight Capacity Exit Index (FCEI), https://research.freightscout.ai

The monthly figures are downloadable as a CSV, one row per state per metric, with the numerator and denominator beside every rate.

The sample floors

A rate built on one or two events is noise wearing a decimal point. A state month is left out unless it clears both floors: at least 5 carriers with a cancellation filed that month, and an insured base of at least 250 carriers. A state that misses is absent. It is not printed as a zero and it does not get an asterisk.

Both numbers are operator set. They were chosen because they are defensible to somebody reading a contract, not because a study of this data picked them. The index holds 60 months of history.

Why the newest month is never the current one

FMCSA filings for a month keep arriving long after that month closes. A recent period is not a quiet market, it is an incomplete one, and publishing it shows a sharp drop that quietly corrects upward later. That is the easiest way there is to mislead somebody here.

So we publish in arrears. A month goes out only once its national count reaches 80 percent of the median of the twelve months before it. Months short of that are computed, held back, and named as held back.

The sample floors cannot do this job. On 9 August 2026 this site was publishing June 2026 with 5.83 percent of its filings arrived. California read 5.39 per 1,000 insured carriers against a range of 84 to 108 in the five months before it, and every cell cleared the sample floor, because 278 carriers clears a floor of 5 easily. A cell floor asks whether a cell is big enough. It cannot ask whether the month has finished arriving.

What we hold back, and why

Anything we compute and do not publish is listed here with the reason. A number that disappears without one looks the same as a number that never worked.

Coverage lapses

Held back since 2026-08-09. A lapse is a carrier that ended the month with no coverage in force having had some during it, so you cannot tell it from an ordinary cancellation until the replacement filing shows up, and replacement filings arrive on a much longer tail. Recent months therefore count reinsured carriers as lapsed.

The lapse count sat between 0.21 and 0.35 of the cancellation count for forty two straight months, then climbed every month to 1.00 at the present edge. Flat for years and then exactly one is missing paperwork, not a market. The arrears gate cannot catch it, because that gate watches cancellation filings and this sits on the other side of the definition.

A state that published before and misses the floor keeps its URL and gets a page saying why, instead of a 404. A citation should not break because one month came in thin.

How to cite this

Published under a Creative Commons Attribution licence. Attribute it and use it for anything, including commercially.

FreightScout, Freight Capacity Exit Index (FCEI), https://research.freightscout.ai

Rates are per 1,000 carriers in the relevant base. A figure quoted without that unit does not mean anything, which is why every rate here carries its unit beside it.