Car-Mart's 82% Collapse Is a Credit Migration, Not a Credit Crunch: Subprime Credit Moved, and $30,000 New Cars Are Now the Tightest Thing on the Lot

In the quarter ended July 31, 2026, America's Car-Mart sold 2,450 retail vehicles, down 81.9% year over year, and revenue fell 57.3% to $145.8 million. A $69.0 million net loss followed, the company said originations were constrained by liquidity, and it consolidated 60 locations during fiscal 2026. This was not a soft quarter. It was a business cutting its own sales by four-fifths.

In the overlapping quarter ended September 30, Credit Acceptance originated $3.46 billion of subprime loans, up 51.0% year over year, and wrote business for 12,093 active dealers, up 18.8%, including 1,350 newly enrolled. One of those numbers counts cars sold from a used-car lot. The other counts credit written to buy them. Read together they say the subprime channel changed hands; they do not say it closed.

The two balance sheets on either side of that transfer explain it. America's Car-Mart had reserved $277.0 million, or 24.74% of net finance receivables, against credit losses at July 31, 2026, and its quarterly net charge-off rate rose to 9.5% of average receivables from 6.6% a year earlier. The captive finance arm of CarMax held a reserve equal to 3.07% of auto loans it financed. One structure is not viable. The other is not stressed.

A buy-here-pay-here lot finances the buyer and keeps the note on its own balance sheet. Federal Reserve researchers found that 81% of loans made to buy-here-pay-here dealers were fully guarantor-backed in the third quarter of 2025, against 62% for other dealers. That structure is why the channel can be switched off in weeks: when the guarantor reprices or withdraws, the lot's product disappears without any change in consumer demand.

America's Car-Mart shows where that leaves a dealer that carried the paper itself. It cut the dealership count from 154 to 94, and in June 2026 amended its lender agreement to support a review of strategic alternatives after covenant breaches. Nine months earlier the fiscal 2026 accounts already showed the direction of travel: revenue of $1.28 billion, down 7.9%, on 48,891 units, down 14.3%.

The gap sits at the bottom of the market. Cars priced at $30,000 or less ran 19 days tighter than the market average at 63 days of supply against 82, while cars priced from $50,000 to $60,000 sat at 91 days. In inventory terms the entry tier turns roughly 30% faster than the market. The average new-vehicle transaction price still reached $50,562 in September, near a record, on a daily sales rate of 34,300 units.

The tightening is not uniform, and September showed the first crack. Cox put entry-price supply at 54 days in August, the tightest of the year, and 63 days in September as the market rebuilt to 82. Entry-price tightness is decelerating rather than accelerating, and that matters for anyone underwriting 2027: the bottom of the new-car market is one credit-availability print away from re-inflating. The Dealertrack Credit Availability Index reached 107.1 in September, its highest reading since April 2015 and its fifth consecutive monthly gain, roughly 9% above a year earlier.

The used-vehicle market is where a repriced subprime portfolio shows up as price. Cox's Manheim wholesale index was 205.9 in September, down 0.6% from a year earlier, and the September used-vehicle report showed supply lengthening by five days as the sales pace fell 6.3%. Certified pre-owned sales were up 10% year over year but down 6% from August. Supply is building while prices grind lower.

The banks that fund this market are getting paid for it. Ally Financial reported third-quarter net interest income of $2.02 billion, up 11.7% year over year, on total revenue of $4.15 billion, with diluted earnings per share of $1.30 against $1.18. Its filing does not separate consumer auto, so treat that as a proxy for bank lending economics rather than proof of an auto-specific gain.

This is not a demand collapse; it is a repricing of collateral. When deep-subprime borrowers are pushed out of used-car financing, their cars do not stop existing — they clear at lower prices to buyers who need no credit. That is a transfer from lenders' balance sheets into used-vehicle prices, and it lands in the gross profit line of whoever owns the retail lot.

The dealer results show the split cleanly. AutoNation reported third-quarter revenue of $7.01 billion, down 0.4%, gross margin of 17.2% against 17.6% a year earlier, and net income fell 24.5% to $162.3 million, with diluted earnings per share of $4.87 versus $5.65. Penske Automotive grew revenue 9.4% to $8.42 billion and lifted diluted earnings per share to $3.36 from $3.23, but gross margin slipped to 15.7% from 16.2%.

Lithia Motors ran the opposite trade: third-quarter revenue of $9.95 billion, up 2.8%, net income of $291 million and diluted earnings per share up 54%, with new-vehicle units up 10.6% and new-vehicle gross profit up 8.5%, which management credited to expanding partnerships with Chinese manufacturers. Nine-month attributable net income of $651.5 million still trailed the prior year's $682.7 million, so the strength is quarterly rather than yet a trend.

The mix consequence is specific: more of 2027's industry gross profit will be earned by people who fund cars rather than sell them. Credit Acceptance and the captive finance arms take the spread, franchised dealers take the volume with a thinner front end, and used retailers keep the units while giving back the price. Expect the 2027 auto market to get bigger and less profitable per car, with the only structural winners being balance sheets that never had to hold a $19,000 note on their own books.

Facts, inference and forecast, separated. The 81.9% unit decline, the $69.0 million net loss, the 24.74% allowance and the going-concern note at America's Car-Mart are facts. Credit Acceptance's 51.0% growth in dollar volume and Cox's 107.1 credit availability index are facts. The claim that subprime credit migrated up the product ladder into entry-price new vehicles is an inference drawn from those facts plus the price-band inventory data, and it should be treated as such: an inference drawn from inventory and volume data, not a measured flow of loans by borrower credit tier.