Why two valuations of the same car disagree

The same model does not sell for the same price everywhere. Compare a house's median for one model with the median for that same model overall, and the differences are large and consistent.

Part of that gap is not the room at all: it is which cars each house sells. A house whose lots for a model are mostly low-mileage, numbers-matching, documented cars will show a higher median than a house selling project cars and drivers, without either being wrong. The table above cannot separate the room from the inventory.

Across all 28 models here, dividing each house's median by the median of the model it is selling — which controls, roughly, for which cars it carries — leaves the houses in this order: Gooding & Company at 188% of the model median across 9 models; RM Sotheby's at 121% of the model median across 18 models; Bonhams at 120% of the model median across 10 models; Barrett-Jackson at 116% of the model median across 16 models; Bring a Trailer at 98% of the model median across 26 models; PCARMARKET at 81% of the model median across 8 models; Cars & Bids at 77% of the model median across 7 models. Read that as a description of the two inventories, not as proof that the same car would fetch more in one room than the other.

A valuation computed last month and one computed two years ago are not the same claim. Markets move, and a model that ran up in 2021 and softened since will read very differently depending on the date attached. Every figure on this site carries the date it was rebuilt, and a stale figure is treated as a defect rather than an inconvenience: the build fails if the stats are more than 45 days old rather than publishing a number that silently aged.

The largest single reason two valuations of "the same car" disagree is that they are not the same car. Engine and gearbox numbers that match the chassis, original paint, a continuous history file, a known ownership chain and an unrestored interior are worth more than the model median on their own. A repaint, a replacement engine and no paperwork move a car the other way.

This is why a model median is a starting point and not an answer. The band around it — the typical range on each model page — is the space in which condition and history do their work.

Ask four things of any number: which model and which year range, how many sales it rests on, what date it was computed, and whether it describes a car or a market. A figure that cannot answer all four is not comparable to one that can, and the disagreement between them tells you nothing.

On this site every model page prints its count and its date in the opening sentence, and its figures are reproducible from the sales it lists. Start at model values, or read the method.