Does a change of operator go with a casualty?

The Coast Guard records inspections and casualties. The Army Corps of Engineers records, each year, who operates each hull. Neither agency joins to the other. Maury does, on the US official number — 10,725 published hulls carry both a casualty history and an eleven-year operator history — so a question neither record can answer alone becomes a count: do hulls that change hands more often appear in more casualty investigations?

Lift is the band’s rate divided by the rate across all 42,580 vessel-years (24.2%). Hulls that changed operator two or more times run 1.51× the rate of hulls that never did. Where the first finding found a spread of 0.05 across deficiency bands, this variable separates the same population.

Older hulls change hands more often and appear in more investigations, so the table above could be age wearing the shape of operator change. The same count, inside each age band:

Casualty rate per vessel-year; a dash where a cell holds fewer than 150 vessel-years and a rate would be noise. Within the two largest bands the ordering holds. It is not purely age.

An association is not a direction. If a change of operator made a casualty more likely, casualties would cluster after changes. If a casualty made a sale more likely, they would cluster before. Counted on the same hulls:

Close to symmetric. Casualties precede a change of operator at least as often as they follow one. Anyone reading this page as “operator churn causes casualties” is asserting a direction the data does not carry — and so would anyone reading it the other way.

It says nothing about any operator. The operator here is a variable — changed, or did not — never a subject. No company is named, counted, or compared on this page, and Maury publishes no statistic for a named operator (why).

A change of filed operator is not proof of a sale. The Corps records who files the return. A change can be a restructuring, a bareboat charter, or a change in who files, with no transfer of the asset. This page counts changes in the filed record and calls them nothing else.

It says nothing about any individual vessel. This is a statement about a population. Maury does not score, rate, or grade vessels.

The window is 2014–2021 — where the casualty cohort and the operator panel overlap — not the archive’s span.

The 2019 return lists vessels but not operator names. For hulls absent from it, a change between 2018 and 2020 is counted once and may be two — 141 of 7,874 changes. Counts are a floor.

Same-company id changes are excluded. Operators can hold more than one filing id; runs whose names normalise equal are one operator. 535 such non-changes were removed before counting.

The cohort is casualty-enriched, as the first finding’s is: rates here are within-cohort and are not fleet casualty rates. The comparison between bands is unaffected.

This is a correlation over a window, not a causal claim.