DGActual Market Brief — Week of August 31, 2026

The number we have been waiting on all summer finally posted. The Fed's Q2 bank delinquency rate came in at 2.62 percent, a touch better than Q1, and it moved our own model up a notch for the second week running. Meanwhile two labor standoffs that could have taken real metal off lots this fall both stood down: Hyundai and Kia workers reached tentative wage deals, and the strikes planned for this week are off the table pending a vote. Add in a Ford gigacasting bet on a sub-$30,000 EV, a Polestar ban that blindsided its own dealers, and an eighth straight week of used retail softening, and this is a week where the good news and the grind both showed up on the same page.

Overview: the delinquency number finally landed, and it was fine

For three straight briefs we told you the Fed's Q2 bank delinquency print was the next real input to our SAAR model, and that it had not posted yet. It posted this week: 2.62 percent, down slightly from Q1's 2.64. Not a dramatic move, but a real one, and it nudged our Q3 base up to 16.08 million. Small number, right direction, no surprises. Sometimes the best data is boring data.

Bigger news on the labor front. Hyundai and Kia both reached tentative wage agreements with their South Korean unions this week, and Kia's planned three-day strike is now on hold pending a ratification vote. If it holds, that is real allocation relief for two brands that spent August thin. Stellantis also restarted Fiat 500 Hybrid production after a 31-day shutdown, though the ramp is slow: about 180 units per shift on day one, building toward 400 a day.

The parts of the market that are not fine stayed not fine. Used retail prices fell for an eighth straight week. And two OEM stories landed hard: Honda indefinitely suspended a $15 billion Canadian EV plant citing U.S. tariffs, and Polestar dealers say they were blindsided by a federal ruling blocking the brand's U.S. sales starting with the 2027 model year. Below, section by section.

Sales and demand: our model ticked up again, this time on credit

The DGActual model's Q3 base moved from 16.07 to 16.08 million this week, a small lift, but it is the second consecutive week the model has improved rather than held or fallen. Two weeks ago it was the sales pace. This week it is credit: Q2 delinquency at 2.62 percent came in better than the 2.64 the model had been running on, and every basis point of improvement there is a basis point less headwind on the forecast.

One input is still not fully resolved. Q3 2026 sentiment remains pinned at 50.0 in our model because a complete-quarter University of Michigan reading is not yet available, a limitation we disclose every time it applies rather than pretend the number is live. J.D. Power's own August forecast lands at 16.4 million SAAR with retail sales down 6.9 percent year over year, hybrid share at 18.2 percent, and the average monthly payment projected at a record $812. NADA's July read: 16.3 million SAAR, $808 average payment, also a record, and incentives up 8.1 percent to $3,451 per unit. Read those together and the picture is consistent: volume is holding up because incentives and rate relief are propping up an affordability picture that keeps setting records for the wrong reason.

Credit: Q2 finally posted, and one regional numbers looks much worse

The headline bank number, 2.62 percent serious delinquency on consumer loans, is mild. A regional read from the Philadelphia Fed is not: their Q2 data shows severe auto-loan delinquency above 9 percent, with serious-delinquency transitions at 3 percent, the highest rate since 2010. That is a sharper read than the national bank number, and it is consistent with the record-payment story running through every forecast this week. National averages can hide regional stress, and this is exactly the kind of gap our model cannot see because it runs on national inputs.

Labor and production: two strikes stood down, one plant restarted slow

Hyundai and Kia both reached tentative wage agreements this week: Hyundai's base pay increase lands at 4.1 percent, and Kia's planned August 26 through 28 strikes are on hold pending a ratification vote. If ratified, this closes out the labor uncertainty that has been thinning both brands' allocations since July, though a tentative deal is not a done deal until members vote.

Stellantis restarted Fiat 500 Hybrid production at Mirafiori after 31 days down, targeting roughly 180 vehicles per shift at restart and about 400 a day at full ramp. That is a real but gradual recovery, not a snap back to normal volume.

Longer-horizon moves worth tracking: Hyundai laid out plans for more than 100 global vehicle launches by 2030, including 58 in North America and 500,000 units of added North American capacity, with 10 new hybrid and extended-range EV models in the mix. Honda said an eighth North American assembly plant may depend on whether USMCA gets extended, with a decision expected within one to two years and a target date around 2030. GM Canada announced roughly 359 million Canadian dollars for Oshawa and St. Catharines, including next-generation Sierra production. And Honda and Nissan are reportedly close to a shared vehicle software and onboard-computer agreement, with joint vehicles potentially arriving as early as 2029.

Trade and tariffs: Honda pulled the plug on a $15 billion EV plant

The biggest tariff casualty of the week: Honda indefinitely suspended its $15 billion Alliston, Ontario EV complex, citing U.S. tariffs directly as the reason. That is not a delay announcement, it is a suspension, and it is the clearest signal yet that tariff uncertainty is now driving capital allocation decisions at OEM scale, not just pricing decisions at the parts-counter level. Volkswagen's restructuring saga also escalated: fresh stakeholder opposition and alternative proposals surfaced ahead of a September 4 board meeting, with up to 50,000 additional job cuts still on the table, and reporting this week showed labor costs at two of VW's EV sites running more than double comparable locations.

EV makers: margins are getting squeezed from Shenzhen to Fremont

A rough week for EV manufacturer earnings. Li Auto's Q2 gross profit fell 53.3 percent year over year, with vehicle margin dropping to 9.4 percent from 19.4 a year ago. XPeng's Q2 gross margin held at 20.7 percent, but Q3 revenue guidance of roughly 21.7 to 23.4 billion yuan landed well below the 26.6 billion consensus. BYD's Q2 profit rose 30 percent year over year but missed a 48 percent forecast, and first-half profit actually fell 21 percent.

On the product side: Tesla raised Cybertruck prices by $5,000 on two trims, dual-motor now $74,990 and all-wheel-drive $84,990, a move that cuts against every other signal in this market pointing toward affordability. And Ford installed 9,000-ton gigacasting equipment as part of a $2 billion Louisville renovation to build the Fathom, its midsize electric pickup officially named earlier this month at $28,350 to start, $29,945 with destination, with preorders opening in early 2027. That is the opposite bet: cheaper, not pricier, and a genuine test of whether gigacasting can get EV manufacturing costs down enough to hold that sticker.

The one that will actually show up on your lot: Polestar dealers say they were blindsided by a federal ruling blocking the brand's future U.S. sales starting with the 2027 model year. If you carry Polestar inventory or have customers in the pipeline, that is a conversation to have proactively rather than reactively.

Safety desk: a full week of recalls, none of them small

Ford Explorer: 6 vehicles for a connecting-rod-bearing seizure risk, carrying do-not-drive and park-outside guidance, the smallest count of the week but the most severe guidance. Lucid Air: 27,185 vehicles under campaign 26V540. Kia EV9: 21,290 vehicles under campaign 26V545. Dodge Durango: 74,578 vehicles across 2021 through 2023 model years. Toyota Tacoma: 48,280 vehicles across 2024 and 2025 model years. And a BMW-family campaign spanning Mini, Rolls-Royce, and multiple BMW lines totaling 428 vehicles across a wide range of trims and years.

None of these are single-brand mega-campaigns like the Toyota Camry Hybrid recall two weeks ago, but taken together it is a genuinely full recall week across seven different nameplates. If you carry any of these brands, check your open-campaign list before Monday's first customer call.

DARPI: eighth straight week down, spread at another record

DARPI is the DGActual Automotive Retail Price Index: used retail asking prices from dealer listings across 11 segments, tracked weekly, indexed to 100 at our June 8 baseline, with the wholesale benchmark rebased to the same 100 so the two lines compare directly.

Week 15: retail fell another 0.29 points to 96.55, the eighth consecutive weekly decline. The retail-wholesale spread widened again to negative 2.76, another record. Midsize Car led the decline this week at minus 1.29, a reversal from two weeks ago when it was one of the segments actually rising. Fullsize Car keeps bucking the trend, up 0.48 and now the most consistent gainer in the index. EV continued lower, down another 0.34, its sixth decline in a row.

On EVs specifically, and this ties directly to Fathom and the Cybertruck price hike above: based on dealer listings, the median near-new EV asking price is now $31,846 at dealerships against $34,000 in FSBO listings from private sellers, with dealer listings turning in about 27 days at the current lot against 79 for private listings. That private premium has held for two straight weeks now, and it keeps looking like the same story: sellers stuck above their loan payoff, waiting for a market that keeps moving away from them.

Purchase picks: what we would buy this week, segment by segment

Every pick below is based on dealer listings only, covering 2022 through 2025 model years, all in the same 20,000 to 60,000 mile band so turn speed and price are compared fairly across nameplates. Days shown are days at the current listing, meaning how long the car has sat on its current lot, not its lifetime on the market.

Compact SUV: 2024 and 2025 Mazda CX-5. The 2025 is the star of the whole board: $27,139 median, turning in 18 days, on the deepest sample of any CX-5 year. The 2024 runs $26,758 at 24 days. Skip the 2023, which turns slowest of the four years at 29 days for nearly the same money. CR-V and RAV4 both run $5,000 to $6,000 higher across the same years and turn slower.

Midsize sedan: 2023 and 2024 Honda Accord. Turn speed is flat across all four Accord years at about 31 days, so buy the year with the best price-to-miles mix: the 2024 at $26,804 on 34,000 median miles, or the 2023 at $26,143 on 36,000. Camry and Sonata turn slightly faster, 26 to 27 days, but both carry roughly 42,000 median miles against the Accord's mid-30s. Fresher cars, comparable clock.

Full pickup: hold off, all years. Every truck we priced is turning slower than any other segment we checked, F-150 at about 48 days and Silverado at about 43 at the current listing, consistent with Full Pickup being the slowest-turning segment in the index at 123 days lifetime. Buy selectively and cheap, not confidently.

EV: 2023 Kia EV6 and 2023 Tesla Model 3, not the category. The 2023 EV6 is the fastest-turning vehicle we measured anywhere this week: 16 days at $27,500 on a real sample of 430 dealer listings. The 2023 Model 3 runs $27,995 at 23 days; note that 2024 and newer Model 3s carry the refreshed hardware and a roughly $7,000 premium, $34,995 and up, so the value year and the newest year are different products. Mach-E and Ioniq 5 both turn slower at similar or higher prices. Same lesson as last week: EV buying strategy is year and nameplate specific, not category-level.

Compact car: 2025 Toyota Corolla, with a 2024 alternative. The 2025 turns in 19 days at $23,021 on 32,000 median miles. The 2024 is cheaper at $21,593 and still turns in 22 days, but carries nearly 49,000 median miles, so it is a payment-buyer car, not a condition-buyer car. Either year beats the Civic on price, $22,000s against the Civic's $25,350, at comparable turn speed.

FSBO acquisition desk: private-seller prices vs dealer prices, mileage-matched

This section is about FSBO listings: for sale by owner, meaning private sellers listing their own cars, not dealerships. The question it answers is simple: for the same model, 2022 through 2025 model years, at comparable mileage, is a private seller asking less than a dealer is asking? When the answer is yes, that is a car you may be able to buy from a driveway for less than you would pay at auction or from a wholesaler. Fresh pull this weekend. Two signals moved enough to be worth flagging on their own.

Here is the full board, this week against last week. Every figure is the gap between what private sellers ask and what dealers ask for the same model at comparable mileage, 2022 through 2025 model years combined. Negative means the private seller is cheaper. The pattern this week: the real private-seller discounts are narrowing, and the models where private sellers overprice, trucks and the Grand Cherokee, are getting worse. Tesla Model Y is the exception worth watching, and its sample is deep enough to slice by year: the discount concentrates in older cars, $1,572 on 2022s, $840 on 2023s, and $503 on 2024s, all at matched 20,000 to 60,000 mile inventory. If you are buying Model Ys out of driveways, the 2022 is where the money is. The F-150 slices the other way: private sellers ask $3,471 over dealer prices on 2022s, $6,861 over on 2023s, and the 2024 private population is so overpriced and trim-skewed that it is not even worth engaging.

Negative = the private seller (FSBO) is asking less than dealers for the same model at comparable mileage. All figures compare 2022 through 2025 model year vehicles in the 20,000 to 60,000 mile band on both sides. n = number of matched private listings.

Honda CR-V held up: private sellers are asking $2,322 less than dealers before adjusting for mileage, and $2,947 less when both sides are matched to comparable miles, now on a sample of 10 matched private listings. Last week we called this promising but unproven on a thin sample. This week it is proven. Nissan Rogue and Tesla Model Y both continue to check out as real buying opportunities: private asking prices run $1,036 and $1,323 below dealer asking prices respectively at matched mileage.

Toyota Camry flipped the other way. Last week private Camrys looked cheaper than dealer Camrys; this week, with a larger matched sample of 16 private listings, they are actually asking $1,273 more than dealers once mileage is matched. Dealer Camrys are running noticeably higher miles than private ones right now, and that mileage gap, not a real price advantage, was creating the illusion of a discount. Toyota RAV4 and Honda Accord remain mileage mix: private sellers look cheaper at first glance, but the gap vanishes once you compare cars with similar miles.

Still anchored regardless of mileage, meaning private sellers are asking more than dealers and will not budge: Ford F-150 private listings run $4,780 above dealer asking prices even at matched mileage, and Jeep Grand Cherokee runs $4,008 above. Both gaps are worse once mileage is accounted for. Patience files: do not chase these driveways.

Full segment data and model-level breakdowns at data.dgactual.com. Dealer reputation across 17,814 stores at reviews.dgactual.com.

Sources: Federal Reserve G.19 Consumer Credit and DRCLACBS. Federal Reserve Bank of Philadelphia Consumer Credit Explorer. J.D. Power and GlobalData U.S. Automotive Forecast, August 2026. NADA Market Beat, July 2026. Reuters (Kia wage deal, Hyundai wage deal and capacity plans, Honda Alliston suspension, Volkswagen restructuring, Tesla Cybertruck pricing, Honda-Nissan software agreement, Honda eighth plant, Ford Fathom naming and pricing). MarkLines (Stellantis Mirafiori restart). Repairer Driven News (Ford gigacasting). The Wall Street Journal (Polestar). XPeng and Li Auto investor relations. NHTSA recall filings 26V532, 26V540, 26V545, 26V528000, 26V514000, and the BMW-family campaign. DGActual Retail Price Index, SAAR model, and FSBO acquisition pull (MarketCheck API). DGActual Market Intelligence. Week of August 31, 2026.