Every dollar figure on a flight row is one of two things: a real amount from a record we hold, labeled Documented cost, or a modeled estimate of air hours times an hourly rate, shown as a range. This page covers where the rates come from, how wrong they can be, and where we refuse to put a number at all.
Where the rates come from
584 priced public-record lines totaling $4.35M: state aeronautics billing logs, charter invoices, foundation reimbursements, payment postings, and booster gift valuations. Those are not all interchangeable. Charter estimates are fit only from paid invoices and reimbursements that can be tied to a complete flight outing. State billing tests owned-aircraft costs; gift valuations and payment postings remain source evidence but do not train charter rates. Where we hold the actual invoice for a trip, the row shows that amount tagged Documented cost and the estimate steps aside. How we get the records: how we source flight records.
Owned, chartered, booster
For charter, we first look for paid outings from the same operator and aircraft type. If none exist, we widen the comparison in steps: aircraft type, operator and aircraft class, then aircraft class. Only after those receipt cohorts fail do we use published market rates. Receipt totals are already all-in—including fuel—so we do not add fuel or federal excise tax a second time. The estimated fuel usage shown on a flight report is separate operating context, not another charge added to the dollar estimate. Older payments are normalized to the flight date with the same aviation-cost index used by the fallback model. A university-owned aircraft has no bill, so we compute two numbers and show both: marginal (fuel, maintenance, reserves: what the flight added) and fully allocated (marginal plus crew, hangar, insurance, and depreciation spread across hours flown, the figure a university’s own budget reports, and the one the headline uses). A booster’s airplane is neither: its rows show an estimated value of the flight as gift-in-kind context, and that value is never added to a program’s spend.
How wrong it is
We refit the model on state billing records dated before 2024, then scored it on 2024 legs it had never seen. The published band contained the real billed amount for 85.4% of those flights (129 of 151); the median midpoint missed by 22.6%, so trust the band, not its midpoint. That holdout is turboprop state-pool billing from one state: a real test of the method, not a measurement of jet accuracy. The best owned-fleet anchor is Tennessee, which publishes its own totals: $3.29M across 647.3 hours, $5,084/hour fully allocated. The Tennessee-specific row is anchored to that published total; it is a reconciliation check, not independent validation.
Because that validation covers turboprops only, the jet classes are banded a different way: from the spread of published per-type operating costs within each class, rather than from the turboprop billing spread. It is a defensible width, not a measured one. A jet program’s band therefore reflects how much published costs disagree for that class of airplane, and only a school that publishes its own flight-operations totals, as Tennessee does, gets a band anchored to real spending.
Charter coverage is still uneven. Some aircraft types have one paid comparable; others have several, and many still fall back to class-wide or published market rates. Every flight report names the basis and comparable count. A one-comparable estimate receives a wider range because it is useful direct evidence, not a mature market sample.
What these numbers do not claim
- Nothing about who was aboard or why. Cost is arithmetic on hours and rates, never evidence of purpose or passengers; those claims are governed by our editorial policy.
- No annual total is ever split into a flight list. A vendor’s yearly total from a tax filing sanity-checks an aggregate; it is never divided across trips we cannot tie to it.
- Some attribution is provisional. Trips matched through an aircraft’s present-day owner say so in their breakdown and will shift as effective-dated ownership fills in.
- An unpriced flight shows nothing. No established aircraft class or usable duration, no dollar figure.
If you hold a record that contradicts a figure here, we want it: see corrections.