Car Repossession Rates in 2026: What the Numbers Measure and What a Repo Recovers

Dealer Playbook · United States

By TradeBasis · Published October 3, 2026 · Figures verified October 3, 2026

---

Car repossession rates have no official national measure. The figure everyone quotes — roughly 1.73 million vehicles in 2024, the most since 2009 — is a Cox Automotive estimate modelled largely from auction throughput, and it structurally excludes the buy here pay here segment, which recovers and retails its own units without ever crossing an auction block. Meanwhile the number that actually determines what a repossession costs, the recovery rate, fell to 32.64% in Fitch’s subprime auto ABS index in December 2025 against a pre-pandemic average of 43.73%.

Measure	Value and period	Source
Estimated US repossessions	1.73 million, 2024	Cox Automotive
Recovery rate, subprime auto ABS	32.64%, December 2025	Fitch Ratings
Pre-pandemic average recovery rate	43.73%	Fitch Ratings
60+ day delinquency, subprime auto ABS	6.90%, January 2026 (record)	Fitch Ratings
Auto loan balances 90+ days delinquent	5.49%, Q2 2026 (2nd highest since 2003)	New York Fed
BHPH average default rate	36.04%, 2018-2021	Subprime Analytics

Credit performance figures are compiled in Table 1C of the TradeBasis BHPH Economics and Portfolio Benchmarks 2026 report, which carries the full population definitions and caveats.

What car repossession rates actually measure — and what nobody measures at all

Ask how many cars are repossessed in a year and you will get a confident answer from almost any source. Start instead with the fact that makes every other number in this article conditional: no agency in the United States collects repossession data. There is no federal register of recoveries, no state filing requirement that aggregates nationally, and no industry body that counts them. Every figure you have seen quoted is a model built on a proxy.

That is not a reason to dismiss the numbers. It is a reason to know which proxy produced the one you are about to cite, because the proxies produce answers that differ by more than a million vehicles. Published car repossession rates are estimates of an uncounted quantity, and the estimate you pick should depend on what you are using it for.

The three bases car repossession rates are built on

Basis	What it counts	2024-2025 figure	What it misses
Auction throughput	Repossessed units crossing a major auction block, scaled by delinquency roll rates	1.73 million (2024, Cox Automotive)	Units never sent to auction: BHPH re-retails, direct disposals, redemptions
Industry reconstruction	Completed recoveries estimated from auction data plus delinquency patterns	1.88 million (2024, CURepossession)	Same blind spots, different adjustments and assumptions
Recovery assignments	Instructions issued to recovery agents, via the Recovery Database Network	10.5 million assignments projected (2025)	Nothing — it over-counts. One vehicle generates many assignments

The third row is where most public confusion originates. An assignment is an order, not an outcome. A lender instructs a recovery agent to locate and secure a vehicle. The borrower cures, so the assignment is cancelled. Three months later the account rolls again and a new assignment issues, possibly to a different agent, possibly to a forwarder who assigns it onward. The vehicle may never be found at all. Treating 10.5 million assignments as 10.5 million repossessions overstates the real figure by a factor of roughly three, which is why the same analysts who publish the assignment data estimate about 3 million actual recoveries from it.

Between the two completed-recovery estimates — 1.73 million and 1.88 million for the same year — the gap is 8.7%. That is a reasonable spread for two models of an uncounted quantity, and either is defensible if you say which one you used.

Why buy here pay here recoveries are missing from the count

Here is the structural blind spot in every estimate of how many cars are repossessed, and it matters more than the 8.7% spread between the two completed-recovery models.

Estimates built on auction throughput can only see vehicles that reach an auction. A conventional lender repossesses collateral it has no retail channel for, so it consigns the unit to a wholesale auction and takes what the lane gives it. That unit is counted.

A buy here pay here operator is in a different position entirely. It originated the contract, it holds the paper, it recovers the vehicle itself or through a local agent, and — critically — it has a retail lot to put the unit back on. Many BHPH recoveries are reconditioned and re-retailed from the same location they were sold from the first time. The vehicle never crosses an auction block, never appears in auction throughput data, and never enters the model that produces the national estimate.

The consequence is uncomfortable: the segment with the highest documented default rate is the segment least visible in the published statistics. Subprime Analytics data for 2018 to 2021, reported in NIADA’s 2021 composite, put the average BHPH default rate at 36.04% with an average net loss per charge-off of $6,365. Senator Warren’s February 2026 letter to NIADA cited BHPH lenders writing off more than 35% of their loans between 2018 and 2020 — an independent figure landing in the same place.

So when a national headline reports that car repossession rates have reached their highest level since 2009, the honest reading is that auction-visible repossessions have. What happened on several thousand independent lots is not in the number, in either direction. Those figures and their population definitions are set out in the credit performance section of the benchmarks report.

Car repossession rates are the wrong number to watch. This is the right one.

Volume tells you how many units came back. It tells you nothing about what came back with them.

The measure that does is the repossession recovery rate: net disposal proceeds expressed as a percentage of the balance outstanding at charge-off. Fitch Ratings publishes it monthly for its subprime auto ABS index, and it has not been near its own historical norm in years.

Period	Recovery rate	Gap to pre-pandemic average
Pre-pandemic average	43.73%	—
July 2023	45.50%	+1.77 pts
February 2024	35.82%	-7.91 pts
December 2024	31.84%	-11.89 pts
December 2025	32.64%	-11.09 pts
March 2026	37.48%	-6.25 pts
Trailing 12 months to March 2026	37.02%	-6.71 pts

Fitch Ratings subprime auto ABS index. The trailing twelve-month average fell from 37.92% a year earlier. Compiled in Table 7A.

Two readings in that table need handling with care. Recovery rates are strongly seasonal. They improve from January through April as tax refunds lift both cure rates and retail demand for cheap used vehicles, and they deteriorate into December. Comparing March 2026 to December 2025 and concluding that recoveries improved 4.84 points is a seasonal artefact, not a trend. The like-for-like comparison is December to December: 31.84% to 32.64%, an improvement of 0.80 points from a very low base.

The trailing twelve-month average is the cleaner signal, and it went the other way: 37.02% against 37.92% a year earlier. Recovery performance is still deteriorating on an annual basis.

What the repossession recovery rate gap costs in dollars

Loss severity is the mirror of the recovery rate. At the pre-pandemic 43.73%, a charged-off account lost 56.27 cents on the dollar. At December 2025’s 32.64%, it lost 67.36 cents. In relative terms loss severity rose 19.7%.

Put a number on it. Take an account with a $9,800 payoff balance at charge-off:

Recovery rate applied	Net recovered	Loss on the account
43.73% (pre-pandemic average)	$4,286	$5,514
37.48% (March 2026)	$3,673	$6,127
32.64% (December 2025)	$3,199	$6,601
31.84% (December 2024)	$3,120	$6,680

Illustrative, on a single hypothetical $9,800 payoff balance. Index recovery rates describe securitized subprime pools disposed of at wholesale; they are not a forecast for any individual account.

The December 2025 rate costs $1,087 more per charged-off unit than the pre-pandemic average would have. A lot charging off forty units a year is carrying roughly $43,500 of additional annual loss purely from the collateral recovering less — before a single underwriting decision is examined.

The repossession recovery rate is a valuation outcome, not a collections outcome

This is the part that gets misdiagnosed, and it is the reason the whole topic belongs to the appraisal desk rather than the collections floor.

Rewrite the definition. Recovery rate is:

net proceeds from the vehicle ÷ balance owed on the account

The denominator was set years ago, at origination, by what you decided the vehicle was worth and how much you were prepared to finance against it. The numerator is set today, by what that vehicle is actually worth in your market in the condition it came back.

Nothing the collections department does changes either term much. A better skip-trace gets the unit back sooner, which helps — sooner means less depreciation and less accrued balance — but it does not change the ratio’s structure. The recovery rate collapsed because the denominator was written against 2021 and 2022 vehicle values and the numerator reverted to a normal market.

Fitch has said as much directly, attributing weaker recovery performance in part to higher ABS balances from weaker-performing 2022 vintage loans originated when used-car prices peaked. That is a valuation statement wearing credit-analysis clothing.

And the backdrop is still softening

Recovery assumptions written six months ago are already stale. The Manheim Used Vehicle Value Index read 206.2 in mid-September 2026, down 1% from August and 0.4% below the same point in 2025. Wholesale supply stood at 27.8 days, up 2.4 days year over year. Cox Automotive’s Jonathan Gregory described the market as having “fully given back the stronger pricing trend from the spring’s tax-refund bounce.”

A falling wholesale index with rising days’ supply is precisely the condition under which recovery rates deteriorate: more units competing for the same lane demand, at lower money. For the wider economic setting those contracts sit in, the TradeBasis USA BHPH Macro Conditions Index read 40.4 out of 100 in August 2026, down 14.2 points from May. The index is experimental, and it describes the national economic backdrop — employment, household budgets and funding conditions — rather than measuring or predicting BHPH loan performance.

How to calculate the true net loss on one repossessed unit

Because the index describes securitized pools disposed of at wholesale, it will not tell you what your own recoveries look like. Six steps will.

1. Start from the payoff balance on the recovery date

Not the original amount financed. Recovery rate is measured against what is still owed, including accrued interest and permitted fees, on the day the unit came back.

2. Value the vehicle in the condition it came back in

A repossessed unit is not a trade-in. Assume missing keys, missing service records, deferred maintenance and cosmetic damage until inspection says otherwise. Value the actual trim and mileage against current market comparables, not against the number you used at origination.

3. Subtract the cost of getting it back and sale-ready

Agent fee, transport, storage, key replacement, personal property handling, reconditioning. These come off before anything is recovered against the balance, and they are frequently the difference between a thin loss and a bad one.

4. Subtract the disposal cost for the channel you will genuinely use

Wholesale carries auction fees, transport and the wholesale-to-retail discount. Re-retailing on your own lot avoids those but adds holding cost for every day it sits. Price the channel you will actually use, not the one that flatters the number.

5. Divide net proceeds by the payoff balance

That is your recovery rate on that unit. The remainder is loss severity. $3,199 net against a $9,800 payoff is a 32.64% recovery and a 67.36% loss.

6. Compare to your underwriting assumption, not to the index

The useful benchmark is the recovery rate you assumed when you wrote the deal. If realised recoveries sit persistently below it, the error is upstream of collections — it is in the valuation at origination. Run twenty closed accounts through this and you will know within an afternoon whether your advance policy is calibrated to the market you are actually in.

The delinquency pipeline behind car repossession rates

Repossessions are a lagging output. The leading indicators are delinquency rolls, and they are at or near records.

Measure	Latest	Year earlier	Source
Subprime auto ABS 60+ day delinquency	6.90% (Jan 2026)	6.45% (Jan 2025)	Fitch Ratings
Subprime auto ABS annualised net loss	9.81% (Jan 2026)	9.51% (Jan 2025)	Fitch Ratings
Auto loan balances 90+ days delinquent	5.49% (Q2 2026)	4.99% (Q2 2025)	New York Fed
Total US auto loan balances	Roughly $1.7 trillion (Q2 2026)	+$28 billion in the quarter	New York Fed

The January 2026 subprime 60+ reading of 6.90% is a record for Fitch’s index, against a pandemic low of 2.58% in May 2021. Delinquency series are compiled in Table 8A.

Note that both measures moved almost identically. The all-borrower delinquency rate at the New York Fed rose 0.50 points in a year, to 5.49%; subprime ABS delinquency rose 0.45 points, to a record 6.90%. This is not a subprime-only problem. The New York Fed’s measure covers every auto loan in the country, prime included, and at 5.49% it sits just below the 5.60% high set in Q1 2026 — the two highest readings in 94 quarters of data. Subprime is worse in absolute terms, as it always is, but it is not deteriorating faster than everything else. The whole book is.

Washington has started asking for the data

On February 4, 2026, Senator Elizabeth Warren, ranking member of the Senate Committee on Banking, Housing, and Urban Affairs, wrote to the chairman of the National Independent Automobile Dealers Association requesting information from its buy here pay here members, with a February 16 response deadline. The letter asked for:

• Annual repossession rates for 2022 through 2025, broken down by credit score, loan amount and time to repossession
• How the twenty largest members conduct repossessions — in house, through outside agents, or through forwarders
• Incidents in which the wrong vehicle was repossessed through member or third-party error
• Consumer complaint volumes alleging repossession errors, dispute timelines and outcomes
• Training materials, manuals and guidance on preventing errors and avoiding breach of peace
• Use of GPS locators, kill switches, starter-interrupt devices and payment-reminder systems

A congressional information request is not an enforcement action and carries no legal obligation in the way a civil investigative demand does. But the shape of the request is informative: it asks for exactly the records a well-run operation already keeps and a poorly-run one does not. The first item — car repossession rates broken out by credit score, loan amount and time to repossession — is a portfolio-performance table, and an operator who cannot produce it is not in a position to price risk either.

Read alongside the $700 million Credit Acceptance settlement announced by the New York Attorney General on September 17, 2026 — joined by 39 states and the District of Columbia, and carrying conduct terms on balance collection after repossession — the direction of regulatory attention is not ambiguous. The collections and regulation section of the benchmarks report tracks both.

What this means for your lot

Three things follow, and none of them are about collections.

Know which of the published car repossession rates you are quoting, and to whom. If a lender, an insurer or a reporter asks, the defensible answer is approximately 1.73 million completed recoveries in 2024 on Cox Automotive’s auction-based estimate, with the caveat that no official count exists and that BHPH re-retails are largely outside it. Quoting the 10.5 million assignment figure as repossessions is the single most common error in circulation.

Measure your own recovery rate, by origination month. The index tells you about securitized subprime pools sold at wholesale. It cannot tell you what your units bring back. Twenty closed accounts and the six steps above will, and the result is the only recovery benchmark that should inform your advance policy.

Treat the appraisal as the loss-control step, because it is. Both terms of the recovery ratio run through valuation. What you decided the vehicle was worth at acquisition set the balance; what it is worth at recovery sets the proceeds. An acquisition number set $500 too high costs $500 of gross on every unit that performs, and adds unrecovered principal on every unit that does not — roughly $441 per unit originated in expected cost at a 36.04% default rate and a 32.64% recovery.

That is the honest connection between appraisal accuracy and a lending business, and it is the whole of it: TradeBasis is appraisal and market intelligence software. It is not loan servicing, underwriting, collections, skip-tracing or GPS software, and it has no BHPH-specific features. What it does is tell you what a vehicle is actually worth in your market — on the day you buy it, and on the day it comes back. For how those unit-level losses aggregate into pool returns, see the deal economics section of the report.

Questions dealers ask

How many cars are repossessed each year in the United States?

There is no official national count. The most widely cited figure is Cox Automotive’s estimate of approximately 1.73 million vehicles repossessed in 2024, the highest since 2009. Industry publication CURepossession put 2024 at 1.88 million on a different basis and projected more than 3 million for 2025, working from roughly 10.5 million recovery assignments recorded in the Recovery Database Network. The figures differ because no agency collects repossession data: each estimate is modelled from a different base, such as auction throughput, delinquency roll rates or recovery-agent assignment volume.

Why do car repossession rates and counts disagree so widely?

Because they measure different things. An assignment is an instruction to a recovery agent, and a single vehicle can be assigned, cancelled and reassigned several times, so assignment volume is always far larger than completed recoveries. Estimates built from Manheim auction throughput count only units that cross a major auction block and therefore miss vehicles sold through other channels, redeemed by the borrower before sale, or recovered and retailed by the originating dealer. Cox Automotive has separately estimated that roughly 80% of loans 120 or more days past due end in repossession, which is a roll-rate assumption rather than a count.

What is the recovery rate on a repossessed vehicle?

In Fitch Ratings’ subprime auto ABS index, the recovery rate was 32.64% in December 2025 against 31.84% in December 2024, and 37.48% in March 2026. Fitch’s pre-pandemic average is 43.73%. Recovery rate is net disposal proceeds expressed as a percentage of the balance outstanding at charge-off, so it is strongly seasonal: it peaks around tax-refund season and troughs in December, and readings from different months should not be compared directly.

Why has the recovery rate fallen below its pre-pandemic average?

Principally because the collateral reverted while the balances did not. Loans originated in 2021 and 2022 were written against vehicle values at their peak, so the amount financed reflected a price level the vehicle no longer commands at disposal. Fitch has attributed weaker recovery performance in part to higher ABS balances from weaker-performing 2022 vintage loans originated when used-car prices peaked. The recovery rate is therefore a valuation outcome rather than a collections outcome: it is the ratio between what the vehicle is worth today and what was lent against it years ago.

Are buy here pay here repossessions included in the national figures?

Largely not, when the estimate is built from auction throughput. A buy here pay here operator finances its own paper, recovers its own unit and in many cases retails that unit again from the same lot, so the vehicle never crosses an auction block and never enters the data used to model national counts. This means the segment with the highest documented default rate is the segment least visible in the published numbers. Subprime Analytics data for 2018 to 2021, reported in NIADA’s 2021 composite, put the average BHPH default rate at 36.04% with an average net loss per charge-off of $6,365.

What percentage of auto loans end in repossession?

No agency publishes a national repossession rate, so any single percentage is an estimate. The closest official measures are delinquency figures, and they are at or near records. The Federal Reserve Bank of New York reported 5.49% of auto loan balances 90 or more days delinquent in Q2 2026 against 4.99% a year earlier, on roughly $1.7 trillion of outstanding auto debt. That is the second-highest reading in the 94 quarters since 2003, behind only the 5.60% recorded in Q1 2026. Fitch’s subprime auto ABS 60+ day delinquency index hit a record 6.90% in January 2026. Buy here pay here is a separate population with far higher losses: Subprime Analytics put the average default rate at 36.04% for 2018 to 2021.

Is the federal government investigating repossession practices?

Senator Elizabeth Warren, ranking member of the Senate Committee on Banking, Housing, and Urban Affairs, wrote to the National Independent Automobile Dealers Association on February 4, 2026 requesting data from its buy here pay here members. The letter asked for annual repossession rates for 2022 through 2025 broken down by credit score, loan amount and time to repossession; how the twenty largest members conduct repossessions; incidents in which the wrong vehicle was repossessed; consumer complaint volumes and resolution timelines; training and breach-of-peace guidance; and use of GPS locators, kill switches and starter-interrupt devices. A congressional information request is not an enforcement action, but it indicates where attention is directed.

The Bottom Line

Car repossession rates are the most quoted and least measured statistic in used-vehicle retail. Roughly 1.73 million completed recoveries in 2024 on Cox Automotive’s auction-based estimate is the defensible headline figure, 1.88 million is a defensible alternative on a different model, and the 10.5 million number in wide circulation counts recovery assignments rather than vehicles. Buy here pay here recoveries sit largely outside all of them, because the unit goes back on the same lot it was sold from.

The operationally important number is not the count. It is the recovery rate: 32.64% in December 2025 against a 43.73% pre-pandemic average, with the trailing twelve-month figure still deteriorating at 37.02% against 37.92% a year earlier. That gap is 11.09 points of extra loss severity, about $1,087 on a $9,800 payoff, and roughly $43,500 a year on a lot charging off forty units.

And that gap is a valuation problem. The balance was set by what someone decided the vehicle was worth at origination; the proceeds are set by what it is worth now, in a market where the Manheim index has turned negative year over year and wholesale supply is rising. Collections cannot fix a ratio whose denominator was mispriced three years ago. The appraisal can.

Sources

• Fitch Ratings, subprime auto ABS index — recovery rates (December 2024, December 2025, March 2026, trailing twelve-month averages and pre-pandemic benchmark), 60+ day delinquency and annualised net loss, as reported May 2026. Compiled in Table 7A of the BHPH benchmarks report.
• Cox Automotive — estimated US vehicle repossessions, approximately 1.73 million in 2024, highest since 2009; roll-rate assumption that roughly 80% of loans 120+ days past due end in repossession.
• CURepossession / Recovery Database Network — year-by-year repossession volume series 2008 to 2024 (1.88 million for 2024) and 2025 assignment volume projection of approximately 10.5 million.
• Federal Reserve Bank of New York, Quarterly Report on Household Debt and Credit, Q2 2026 — auto loan balances up $28 billion in the quarter to roughly $1.7 trillion, $211 billion of Q2 2026 originations, and 5.49% of auto loan balances 90 or more days delinquent (percent of balance 90+ days delinquent by loan type) against 4.99% in Q2 2025 and a series-high 5.60% in Q1 2026. Note that this is the stock of seriously delinquent balances, not the separate flow-into-serious-delinquency series, which runs near 3%.
• Cox Automotive, Manheim Used Vehicle Value Index, mid-September 2026 — index at 206.2, down 1% month over month and 0.4% year over year; wholesale supply 27.8 days.
• United States Senate Committee on Banking, Housing, and Urban Affairs — letter from Ranking Member Elizabeth Warren to the National Independent Automobile Dealers Association, February 4, 2026, requesting buy here pay here repossession data.
• NIADA, 2021 Used Car Industry Report composite, and Subprime Analytics 2018-2021 — average default rate of 36.04% and average net loss per charge-off of $6,365. Compiled in Table 8A.
• New York State Attorney General — $700 million multistate settlement with Credit Acceptance Corporation, September 17, 2026.
• Compiled in TradeBasis, BHPH Economics and Portfolio Benchmarks 2026.

Related reading from TradeBasis

Resource	What it covers
What Interest Rate Do Buy Here Pay Here Loans Actually Charge?	The 25.39% derived rate in the Federal Reserve’s matched sample, why it is derived rather than disclosed, and why quoted BHPH rates disagree so widely.
BHPH Economics and Portfolio Benchmarks 2026	The full research report: market size, credit performance, deal economics, funding, collections and regulation, with every figure sourced and dated.
Used Car Holding Costs in 2026	What a day on the lot costs, and why re-retailing a recovered unit is only cheaper than wholesale if it moves.
Negative Equity Trade-Ins in 2026	How lender advance caps decide whether a deal funds, and why ACV accuracy is the cheapest lever in the structure.
FTC CARS Rule Status in 2026	Why the vacated rule is not in force, and what federal and state requirements still apply to dealer pricing and disclosure.
TradeBasis vs vAuto · TradeBasis vs Carbly	Side-by-side comparisons of TradeBasis against the two tools independent dealers most often evaluate.

About TradeBasis

TradeBasis is independent used-vehicle appraisal and market intelligence software for US franchised and independent dealers, built by Basis Software Inc. It values a vehicle against live US dealer listings rather than lagging guide-book averages: comparables are matched by trim, every comparable is shown and linked to its source ad where available, the appraiser can exclude outliers and watch the value recalculate, and the valuation adjusts for mileage and days on market.

Lite is $99 USD per month for one user with unlimited VIN-based appraisals, live US market data, trim-aware comp sets, mileage and velocity adjustments and one-click competitor ad views. Pro is $249 USD per month for up to ten users and adds the full appraisal worksheet, reconditioning cost tracking, printable trade walk-down reports, a Workbench KPI dashboard, inventory management, and appraisal history and search. Both are month-to-month with no contract, no setup fee and all data included. There is no free trial. See plans and pricing.

TradeBasis is not affiliated with Cox Automotive, CDK Global, Fitch Ratings, any OEM or any auction house. It is not loan-servicing, underwriting, collections, skip-tracing or GPS software and it has no BHPH-specific features.

---

Figures in this article are reproduced from the TradeBasis BHPH Economics and Portfolio Benchmarks 2026 report and from the primary sources named above, with their original periods and populations. No official national repossession count exists; all volume figures are third-party estimates and are labelled with the basis used. Index recovery rates describe securitized subprime pools disposed of at wholesale and are not a forecast for any individual account or portfolio. Dollar examples are illustrative. This article is informational and is not legal, tax or financial advice; confirm repossession, notice and deficiency requirements with counsel in each state where you write contracts.