The Salvage Price Index — Commercial Trucks

Measured medians of realized US salvage-auction sale prices for medium/heavy-duty truck lots — what they actually sold for, not estimates. 9,466 realized sales, 2026-06-08 through 2026-08-24. Regenerated weekly from the same pipeline that measures our valuation model. Related: all vehicles · Motorcycles.

Weekly medians across all vehicle lots with a recorded sale outcome. Series begins 2026-06-08; the window extends automatically each week. Copart sale records are collected at effectively full coverage; IAAI collection is partial — an estimated 15–20% of IAAI's sold volume (measured against independent archive counts, July 2026) — so the series is predominantly Copart-weighted. The by-auction-house table shows the measured mix. Medians are unadjusted for lot mix.

Withheld weeks. The following weeks are excluded from this series and all segment figures rather than published from data we know to be incomplete: 2026-06-29 (a collection incident under-sampled low-value lots that week (sub-$600 share 22% vs a 35–40% baseline; n 32,668 vs 44,000–56,000 in adjacent weeks), so its median would overstate the market; withheld pending backfill from native auction records); 2026-07-06 (same collection incident, partially recovered (sub-$600 share 27% vs a 35–40% baseline; n 35,129); withheld pending backfill from native auction records).

Segment cells require at least 50 sales. All figures are medians of recorded sale prices; segment medians reflect the mix of vehicles in that segment (a low median in a damage category can mean cheaper cars, not just deeper damage). The auction feeds spell some manufacturers more than one way (HARLEY-DAVIDSON and HARLEY DAVIDSON, MERCEDES-BENZ and MERCEDES BENZ). From methodology v1.3 the by-make table counts each manufacturer once, under its most common spelling; vintages carrying an earlier methodology version show those marques split across two rows with separate medians. Model names print in the manufacturer's own style where the feed's commonest spelling drops it (F-150, not F150). From methodology v1.4 that style is declared per model rather than derived from the feed; the label changes, the cell does not, and no median moves. Vintages carrying an earlier methodology version show the feed's spelling. From methodology v1.4 the by-make and by-make/model rows carry the measured spread of recorded sale prices beside the median: the middle half (25th-75th percentile) on every cell, and the 5th and 95th percentiles only where the cell holds at least 200 sales. These are percentiles of what sold, not a prediction interval for any one vehicle: a cell mixes every damage type, title brand and odometer reading in it. Medians published under earlier methodology versions are unchanged.

Revisions this week

Salvage sales can take days to settle, so recently published weeks gain late-arriving sales; upstream data corrections are applied when found. We disclose every change to a previously published number rather than restating silently (methodology v1.4; dated snapshots of every published series are archived):

Method, briefly

Residual continuously collects realized sale outcomes from the major US salvage auction platforms — recorded sale results, not asking prices or estimates. This page reports medians of those recorded prices, grouped by week and segment, with the sample size beside every number. The raw underlying listings are not republished. Data is also available programmatically as JSON (methodology v1.4; previously published values are never silently restated — revisions are disclosed above and dated snapshots of every published series are archived).

The same realized-price pipeline trains and evaluates our valuation model — see how we measure accuracy or the API documentation.