Show your work

Data methodology

Every price and purchase statistic on this site comes from the same place: a record of 690,000+ vehicle purchases going back to 2013. Here is how we turn that into the figures you see, and what those figures can and can't tell you.

The data set

  • Source: completed vehicle purchases made by The Clunker Junker, the company that operates Lemondrove and has been buying cars online since 2009. Lemondrove launched in 2026 and its purchases join the same record.
  • Coverage: 2013 through October 2026, all 50 states and D.C. The monthly Lemon Index starts in January 2015; before that there are too few purchases a month to publish one.
  • Size: 695,682 purchases, including 26,895 in the most recent 90 days.
  • What a record holds: year, make, model, state, the seller's answers to the condition questions, title status and the amount paid. Mileage in miles, the offer the seller accepted and a scheduled pickup date are on record only for more recent purchases, and not for all of those.
  • This release: November 17, 2026, covering purchases through October 2026. We plan to update monthly.
  • Other figures: repair cost ranges are rough editorial guides, the scrap index is the U.S. Bureau of Labor Statistics series named below, and state fees come from each state's own pages. None of those are purchase data, and each is labeled where it appears.

Definitions

The Lemon Index
The mean amount paid, by calendar month, for vehicles whose sellers reported that the car does not drive. Nominal U.S. dollars, not adjusted for inflation. State averages and the by-type averages use the same basis.
Typical offer (model pages)
The median amount paid for that model, in any condition, among purchases since 2021. The yearly trend chart on a model page goes back further and shows the median for each year of sale that has at least 20 sales of that model; years with fewer are left off, and the newest year is a part year.
Median offer by model year
The median amount paid for that model year among purchases since 2021: once over every sale of the year, and once for each reported condition. Where a condition has fewer than 5 sales in a model year, no figure is given for it (a dash in the table, an empty cell in the download). Nothing is estimated, smoothed or put in order, so in some model years the cars that didn't run sold for more than the ones that did.
Runs and drives / Doesn't drive, body intact / Doesn't drive, body damage
From the seller's own answers. Runs and drives: the car starts and drives. Doesn't drive, body intact: it does not drive, whether or not the engine starts, and has no major dents or missing panels. Doesn't drive, body damage: it does not drive and has major dents or missing body panels. The Lemon Index covers both kinds of car that doesn't drive.
What shape they were in
The share of cars for which the seller reported each condition: won't start, starts but won't drive, body damage, no title, missing catalytic converter, flood or fire damage. A car can be counted under more than one. We ask about the car's condition, not what failed and not why the owner is selling, so we do not report engine, transmission or head gasket failures as such. Research and model pages count purchases since 2021; the selling pages count the last 90 days, so the two sets of shares differ a little.
Paid as quoted
The share of completed purchases in the last 90 days where the amount paid equaled the amount the seller first accepted, to the dollar. It is counted from the first acceptance on purpose: when a price is changed and the seller agrees to the new one, the sale counts as changed, not as paid in full. The reason for a change is not recorded.
Time to pickup
Calendar days from the day the seller first accepted the offer to the pickup date on the sale, for purchases where both were recorded (about half of them). The records hold a pickup date, not a time of day, so we count days, not hours: zero is the same day, one is the next day. Long waits are counted, in the "6 or more days" share; only a pickup date earlier than the acceptance is left out. The day of acceptance is taken in the main time zone of the seller's state. We report the shares by day and the median. It is the scheduled date, not a record of when the truck arrived.
Changes on last month and last year
The plain difference between two months of the published series, so anyone can reproduce it from the open data. One caution comes with it. A sale is counted in the month its offer was requested, but it only becomes a sale when the car is picked up, days or weeks later, and the cars that take longest tend to be the dearer ones. So the newest month is provisional when we publish it. For each of the last 12 months we compared the Index as it stood 16 days after the month ended (the age of the newest month in this release) with its value now: since then it has moved between 0.5% down and 3.9% up, typically 1.5% up, and a month can keep moving a little after that. The scrap price per ton, checked the same way, moved between 1.1% down and 0.8% up. Where that could reverse a change, we show no arrow and say it is too close to call. We publish no 30-day changes for states, vehicle types or conditions: checked against the figures those periods settled at, they were not reliable enough to print.
Scrap car price per ton
The median amount we paid per ton of curb weight, in a calendar month, for vehicles that do not drive and are at least 20 model years old. It is the price of a whole end-of-life car, not the price of shredded steel at a mill. Months with fewer than 100 such sales are estimated from the U.S. Bureau of Labor Statistics producer price index for iron and steel scrap (series WPU1012), scaled to our nearest measured months.

Minimum sample sizes

We do not publish a figure unless it rests on enough sales to mean something:

  • A model gets its own page only with at least 200 purchases since 2021.
  • A model-year row needs at least 10 purchases and is shown with its sample size, so you can judge it. Rows with fewer than 25 sales should be read as rough guides. A condition within a model year needs at least 5 sales, or its cell is left empty.
  • State figures use the last 90 days where the state had enough purchases in that time, and the last 12 months where it didn't. Each state page says which window it uses, and a figure that still has too few sales behind it is left blank.

Privacy

Published figures are aggregates, with one exception. The "Recently sold" lists show single sales from the two weeks before a data release, reduced to the model year, model, reported condition and a price rounded to the nearest $25. They carry no state, no date and no pickup detail, and they are in no particular order. A model year is only eligible if at least 10 of it were bought across the country in the 90 days before the release and at least 3 in those two weeks, so a row is never the only sale of its model year in the period. We never publish names, contact details, addresses or VINs.

Known limitations

  • It is one buyer's data, not the whole market. These are prices paid through one buying network. Private sales, trade-ins and other buyers may differ.
  • It skews old and broken. The cars in this record are mostly near the end of their lives. The condition figures are not a reliability ranking, and should not be read as one.
  • Condition is self-reported. Sellers describe their own cars, and the answers are published as given.
  • Averages hide spread. The Index is a mean across very different vehicles. Your car's value depends on its specifics.
  • Regional noise. Small states can swing from month to month on a handful of sales.
  • The long-run trend is not only the market. The Index is what we paid, and what we buy has changed. Over the series it has risen far more than the public price index for scrap iron and steel (U.S. Bureau of Labor Statistics series WPU1012). Part of that is the mix of cars and the ways we can resell them, not the price of any one car. Use the Index to compare months within a year or two; be careful comparing 2016 with today.
  • The newest month is provisional. Some cars from the latest month have not been picked up yet when we publish. Its count rises, and its average usually rises a little, in the next release (the range is under "Changes on last month and last year" above).
  • The newest 90 days are provisional too. Sales with a changed price or a slow pickup are the last to complete, so the paid-as-quoted share and the share of pickups scheduled within two days start out a little high. In each of the last 5 periods, as they stood at the same age, they were 0.4 to 0.8 points and 0.7 to 1.0 points higher than they are now.

Using and citing the data

Our published data sets are free to use under the Creative Commons Attribution 4.0 license. Download them from the open data page. Please credit "Lemondrove" and link to the page you used.

Suggested citation: Lemondrove, "The Lemon Index," October 2026 release, lemondrove.com/value/.

Questions about the numbers, or a figure that looks wrong? Email press@lemondrove.com.