Credit Acceptance Corporation (NASDAQ: CACC) — A Disciplined Survivor in a Bad Neighborhood, Now Fully Priced

Company: Credit Acceptance Corporation | NASDAQ: CACC | CIK 0000885550 | FYE Dec 31 | HQ Southfield, Michigan | IPO 1992 Sector: Financials — Consumer Finance / deep-subprime indirect used-auto lending (GICS sub-industry: Consumer Finance) Date: 2026-06-05 | Reference price: ~$544–548/sh (market quote, 2026-06-04/05; public market-data aggregators) | Market cap: ~$5.7–5.8B | Shares: ~10.46–10.68M

⚡ Claude’s Take

This block is the author’s own subjective opinion. It is general information and not investment advice. Everything below it (the analysis, sections 1–15) deliberately carries no recommendation and no price target.

Verdict: HOLD at ~$545. Accumulate-on-weakness below ~$460–480; not a short despite the crowded ~29%-of-float short. Quality owner-operator compounder, re-rated — no longer the bargain it was seven months ago. Framing: quality-compounder-at-a-fair-price with a contrarian tail, not a deep-value entry. The defensible zone is roughly $460–$520 (≈ 10–11× a normalized ~$45–50 EPS, ≈ 3.0–3.4× book); above the high-$500s/~$580 52-week high you are paying the bull case; below ~$460 the earnings-lens margin of safety reappears and the float-shrink machine does the rest.

CACC is a genuinely good business — ~26% ROE, a 30-year per-share-compounding record, a real (if narrowing) edge in deep-subprime loss-forecasting and servicing scale, and the best capital-allocation incentive structure I have seen in consumer finance (an Economic-Profit/EVA comp that charges management for the capital it consumes). The market is pricing it correctly on three things it has gotten right: the FY2025 earnings recovery (NI $247.9M→$423.9M), the documented funding-cost inflection (warehouse spreads cut to SOFR+185, Nov-2025 ABS at ~5.1%), and the regulatory de-risking (CFPB withdrew as plaintiff; the residual NY-AG settlement path is an immaterial ~$75.5M). What it may be mispricing is the durability of the ~26% ROE — which is flattered by ~5.3× leverage and a deliberately shrunk equity denominator, and which management’s own Economic Profit metric says is compressing (it fell every year, $574M in 2021 → $173M in 2025). The single fact that keeps me at HOLD rather than accumulate: subprime market share is still falling (~6.5%→5.1%, and volume-per-dealer −14.4% in the filing) into a record-stretched consumer (subprime 60+ delinquency at a 32-year high ~6.9%). You are being asked to pay ~3.5–4× book — a multiple a clean steady-state model only justifies at a ~30%+ perpetual ROE — for a franchise that is contracting. The ~30% re-rate since the November lows has banked most of the recovery; the asymmetry from here is no longer in your favor.

Conviction: medium. The interesting tension is that the stock is cheap versus its own 10-year history (composite valuation ~37th percentile; P/S ~7th percentile) and ~29% of the float is short — a configuration where an owner-operator buying back ~12% of shares a year can squeeze the bears. Flips bullish if: the 2024 and 2025 loan vintages hold their clean ~22% spreads through the next two 10-Qs and market share stabilizes — proof the underwriting model self-corrected durably and the franchise is no longer shrinking. Flips bearish if: the forecast-change provision re-accelerates (the 2024/2025 vintages start slipping the way the 2022 book did, spread crushed to 11.9%) or share loss accelerates below ~4.5% — that would mean the moat is being competed away, and at ~3.5–4× book the de-rate would hit the multiple and the earnings together. One-line tag: a disciplined survivor in a bad neighborhood — wonderful machine, wrong-ish price.

Standing note on the body (sections 1–15): The analysis below carries no BUY/SELL recommendation and no price target. Valuation is discussed only as embedded expectations and bear/base/bull scenarios. The single position-taking exception in this document is the Claude's Take block above, which is explicitly the author’s own view.

1. Executive Summary

Credit Acceptance Corporation is a deep-subprime, indirect used-auto lender: it does not lend to consumers directly but enrolls used-car dealers as partners and finances the consumers those dealers would otherwise be unable to sell to. Its defining feature is the Portfolio Program, under which CACC advances cash to the dealer against future loan collections and shares the upside via a subordinated “dealer holdback” — so the dealer co-underwrites credit quality and keeps skin in the game. A secondary Purchase Program buys loans outright. The economic engine is a ~50-year proprietary dataset that forecasts collection rates on each loan cohort, recognized as finance-charge income on a level-yield basis; the spread between forecasted collections and the cash advanced is the core profit.

The investment question is a clean binary: is CACC’s >30-year record of ~20–55% ROE and relentless per-share compounding a durable structural moat, or a regulatory-arbitrage business living on borrowed time in a structurally bad, capital-cyclical, politically targeted industry? The honest answer from the evidence is “a genuinely good franchise that is narrowing, in a bad industry that is currently handing it a cyclical tailwind.”

The bull facts are real: FY2025 GAAP net income recovered 71% to $423.9M (from a $247.9M FY2024 trough); Q1-2026 net income rose 28% to $135.8M with the forecast-change provision ($54.4M) at its lowest in years; the 2024 and especially 2025 loan vintages are underwritten cleanly (2025 spread 22.2% vs 22.0% target) after the 2021–2023 over-advance episode; funding cost inflected downward (warehouse spreads cut to SOFR+185, Nov-2025 ABS ~5.1% all-in); the CFPB withdrew as plaintiff and the residual NY-AG matter carries only a ~$75.5M preliminary cash settlement (immaterial to a ~$8.6B-asset company); and a textbook owner-operator buyback machine retired 12.6% of shares in FY2025 alone.

The bear facts are equally real and harder to wave away: subprime market share is still falling (third-party estimate 6.5%→5.1%; the filing corroborates with volume-per-dealer −14.4% and unit volume −12.6%), which fails the central market-share-stability test for a moat; ROE is leverage- and buyback-flattered (~4.9% ROA × ~5.3× equity multiplier, on a shrinking equity base) and management’s own Economic Profit metric fell every year from $574M (2021) to $173M (2025) — true economic returns are compressing even as GAAP recovers; the deep-subprime consumer is maximally stretched (subprime 60+ delinquency at a 32-year high); leverage is ~4.2× and rising as buybacks run at ~3.3× book partly on borrowed money; and a platform-/marketing-CEO (Vinayak Hegde, ex-Airbnb/Amazon, effective 2025-11-13) inherited a quant-underwriting business just as its long-time Chief Analytics Officer retired — key-person continuity risk into the most important part of the model.

On valuation, the apparent puzzle — a premium ~3.5–4× price/book paired with a low ~13.6× trailing / ~10× forward P/E — resolves to: the P/E is the real anchor; the high price/book is largely an artifact of buying back stock at ~3.3× book, not independent evidence of cheapness or a downside floor. A static justified-multiple model says ~3.5× book requires a ~30%+ perpetual ROE; the base-case normalized ~22–26% ROE only justifies ~2.0–2.9× book. So on the asset lens the market is leaning toward the bull case, while the earnings lens is merely fair. The downside is not well cushioned by the low P/E, because a bear outcome (vintage re-slip + continued share loss) would de-rate the multiple and the earnings simultaneously. No price target or recommendation is offered in this body; the scenarios and embedded expectations are laid out in the Valuation section.

2. Business Overview

What it does. CACC is a specialty finance company that enables franchised and independent used-car dealers to sell vehicles to consumers whom traditional (prime/near-prime) lenders decline — the deepest tier of US subprime auto credit, frequently borrowers with no credit score or a damaged thin file. CACC does not originate to consumers directly; the dealer is its customer and distribution channel. (FACT — 10-K FY2025, Item 1.)

Two products, one engine. Volume runs through two contractual structures:

• Portfolio Program (the core, ~74% of FY2025 unit volume). CACC advances cash to the dealer (a “Dealer Loan”) against the future collections of the consumer loan it services. Collections flow through a waterfall: (i) collection costs, (ii) CACC’s servicing fee (~20% of collections), (iii) repayment of CACC’s advance plus an administration fee, and (iv) any residual to the dealer as “Dealer Holdback.” Because the dealer is paid last and subordinated to CACC’s advance recovery, the dealer co-underwrites loan quality — the structure disciplines adverse selection. (FACT — 10-K FY2025, Item 1, collections waterfall.)
• Purchase Program (~26% of FY2025 unit volume / ~28% of dollar volume). CACC buys the consumer loan outright for a single payment and keeps all collections. This is closer to how most subprime competitors operate. (FACT — 10-K FY2025 MD&A.)

FY2025 mix was Portfolio 74.2% / Purchase 25.8% by unit (74.0/26.0 in 2023; 78.7/21.3 in 2024); 71.7/28.3 by dollar; net Dealer Loans were 72.1% of total net Loans receivable. (FACT — 10-K FY2025 MD&A.)

How it makes money. Finance-charge income is recognized on a level-yield basis off a forecasted collection rate assigned to each loan cohort at origination, re-evaluated monthly against actual collections. FY2025: average portfolio yield 26.9%, average net Loans receivable $7,956.3M, finance charges $2,141.8M. The profit driver is the spread between forecasted collections (~64–68% of contractual balance on recent vintages) and the cash advanced/paid to the dealer (an advance rate of ~45–47%). Ancillary income (vehicle-service-contract and GAP administration fees, reinsurance premiums) is minor. (FACT — 10-K FY2025 MD&A; finance charges 92%+ of revenue.)

Revenue composition (FY2025). Total revenue $2,317.2M = finance charges $2,141.8M (92.4%) + premiums earned $95.6M (4.1%) + other income $79.8M (3.4%). Revenue is highly recurring and contractual — it is recognized over the multi-year life of a self-liquidating loan book, not transaction-by-transaction. (FACT — 10-K FY2025.)

Customers and end markets. The direct customer is the dealer: 15,745 active dealers in FY2025 (a record), up from 15,463 (2024) and 14,174 (2023). Concentration is moderate — the top-5 states (MI/TX/OH/NJ/FL) account for 29.2% of dealers, and no single dealer is >10% of receivables. (FACT — 10-K FY2025, Item 1.) The ultimate end market is the deep-subprime used-vehicle buyer — a large, persistent, and economically fragile population.

Scale. ~2,314 employees; ~$8.6B total assets; ~$11.5B gross loans receivable; ~$7.9B net loans. A single-product, single-geography (US), purely organic franchise — no material M&A in its history. (FACT — 10-K FY2025; SEC EDGAR XBRL, CIK 0000885550.)

3. Industry Dynamics

Structure and size. US auto loan + lease balances stand at a record ~$1.68T (Q1 2026; NY Fed/Equifax). Subprime (<620 FICO at origination) is ~15.3% of vehicle-financing dollars and ~22.5% of used-vehicle financing — and subprime share is rising modestly off the 2023–24 trough, i.e., the segment is re-opening, not collapsing. CACC plays only the deepest tier of the used slice, through ~15,700 dealers — a thin SOM within a large TAM. (FACT — Experian State of the Automotive Finance Market, Q4 2025; NY Fed, Q1 2026.)

Competitive intensity — fragmented, no dominant deep-tier firm. The 10-K names the competitive set: buy-here-pay-here dealers, banks, captive finance arms of manufacturers, credit unions, and independent finance companies, “many [of which are] much larger and [have] greater resources than us.” The resolved peer set: Santander Consumer/SC USA (largest subprime), Westlake Financial, Exeter Finance, American Credit Acceptance, Global Lending, OEM captives moving down-market in easy-credit windows, AI-underwriting entrants (Pagaya), plus adjacent OneMain and BHPH lots (America’s Car-Mart) and the Carvana/CarMax finance arms. Banks rebounded used-financing share in Q1 2025 — easy-credit competition cycles in and out. (FACT — 10-K FY2025, Item 1/1A; Experian.)

Industry-structure verdict — structurally a BAD industry. Fragmentation plus the cyclical entry/exit of banks and captives means weak industry-level barriers to entry. Capital is commoditized, the business is brutally credit-cyclical, and it is a permanent political target. The industry does not favor specialists by structure; it favors them only when they out-execute on three firm-specific dimensions: (a) accurate deep-subprime loss forecasting, (b) servicing/collections scale, and © the cheapest, most durable non-recourse term funding through the cycle. The profit pool sits with whoever combines all three — a firm-specific, not industry-specific, pool. (INTERPRETATION.)

Roles in the value chain and where the profit pool actually sits. The chain runs: consumer → used-car dealer (originates the sale, sets the vehicle price and markup) → finance provider (CACC or a rival, who funds the receivable and forecasts/bears credit risk) → capital markets (ABS investors and warehouse banks who fund the provider, largely non-recourse) → servicer/collector (often the provider itself). CACC is unusual in occupying three of these roles simultaneously — risk-bearer, servicer/collector, and, via the dealer holdback, a risk-sharer with the dealer — which is the structural source of whatever edge it has. But the rent in this chain is thin and contested: the dealer captures the vehicle markup (the largest, most reliable slice); the ABS market prices the funding spread to clear (BB tranches ~425–575bps, B ~575–850bps over in 2025 — capital is paid well but is available); and the finance provider is left with the residual collection-minus-advance spread minus funding cost minus servicing cost minus realized losses. That residual is exactly what compressed on CACC’s 2021–2023 vintages (spread crushed to 11.9% on the 2022 book) and what the +178% funding-cost ramp ate into. The structural lesson: the profit pool accrues to the finance provider only in years when its loss-forecasting is accurate and its funding is cheaper than the marginal competitor’s — a conditional, not structural, rent. A 30-year track record matters precisely because it is evidence (not proof) that CACC has captured that conditional rent more often than not. (INTERPRETATION; S&P ABS spread data.)

Capital-cycle read — the central industry call. The deep-subprime tail is in a late-bust / early-recovery phase: marginal and fraudulent capacity is being flushed out while end-demand persists. Tricolor Holdings (a ~$2B-ABS subprime auto lender) filed Chapter 7 on 2025-09-10 and its founder/CEO was federally charged with systematic fraud (~$800M fictitious collateral) on 2025-12-17; Automotive Credit Corp paused all originations (2025-08-07); American Car Center collapsed into Chapter 7 (March 2023). On supply-side logic, capacity withdrawal by weak players plus persistent demand improves competitive conditions and pricing discipline for disciplined survivors. CACC — family-controlled, 30+ years, cheap term funding — is positioned as a survivor, not a forced seller. (FACT — CNN 2025-09-12; CNBC 2025-12-17; Auto Finance News.)

Crucial caveat — this is NOT a clean capital-cycle trough. Total auto ABS hit a record ~$127B in 2025 (second consecutive record; 34% of all ABS issuance); investor demand was “well digested.” Capital has not broadly fled subprime auto — the exits are idiosyncratic (fraud/under-scale) layered on a still-liquid market. The “everyone fled, buy the survivor” thesis is therefore only partly true: easy-credit bank/captive competition can return quickly if losses normalize. (FACT — S&P Global Ratings; Asset Securitization Report, 2025–26.)

Funding bifurcation — the moat made visible. Post-Tricolor, ABS investors demand more collateral verification (“trust but verify”), raising friction for marginal issuers. CACC, by contrast, cut its funding cost in 2025: warehouse spreads dropped to a uniform SOFR+185 (from SOFR+210/+225), and the Nov-2025 $500M ABS priced at ~5.1% all-in to repay higher-cost debt, with ~$2.0B unused capacity retained. Disciplined incumbents get cheaper funding precisely as it dries up for the weak — instability creating stability for the survivor. (FACT — CACC 8-Ks, 2025-07/09/11.)

Collateral dynamics — stressed borrower, recoverable asset. Wholesale used-vehicle values are firm (Manheim index 212.6 in May 2026, +3.6% YoY), cushioning loss severity on repossessions, even as loss frequency peaks: subprime auto ABS 60+ day delinquencies hit a record ~6.9% (Jan 2026; a 32-year high), versus ~0.4% prime — a >10× gap. The deep-subprime consumer is maximally stretched, which is exactly what is killing the weak lenders. (FACT — Cox/Manheim; Fitch; S&P; Wolf Street.)

Regulatory landscape — an asymmetric shift. The federal posture is the friendliest in a decade: the CFPB withdrew as plaintiff from the CACC suit (2025-04-29), dismissed 16 enforcement actions, lost ~100+ enforcement staff, and proposed raising the auto-finance supervision threshold from 10,000 to 1,000,000 loans/yr (cutting supervised auto lenders from 63 to ~5 — CACC almost certainly out of routine federal supervision). But risk migrated to the states: the NY AG remains the sole CACC plaintiff, and on 2025-12-19 NY enacted the FAIR Business Practices Act, importing federal “unfair/abusive” standards into state law, with AG James explicitly intending to “fill gaps” left by the federal retreat. Net: a near-term tailwind with a fatter, less predictable left tail (multi-jurisdiction state-AG risk), not a structural all-clear. (FACT — American Banker; consumerfinance.gov; Morgan Lewis; 10-K Legal Proceedings.)

Industry verdict. Structurally a bad industry (weak barriers, commoditized/cyclical capital, permanent political targeting), currently handing a disciplined survivor a cyclical tailwind (marginal-capacity exit + funding bifurcation + federal-regulatory relief). The two real threats are not the usual oversupply risk but (1) the deep-subprime consumer remaining broken and (2) the regulatory tail migrating to activist state AGs.

4. Competitive Position

Name the mechanism. The genuine, namable advantage is economies of scale + customer captivity in deep-subprime underwriting data and servicing/collections — not the much-cited dealer-holdback structure. A ~50-year proprietary loan-performance dataset feeds a credit-scoring system that forecasts collection rates by cohort and is re-evaluated monthly against actuals; this is a learning-curve/intangible asset a new entrant cannot replicate without decades of loss data through a full cycle. The fixed-cost servicing/collections platform (predictive dialer, repossession/remarketing, deficiency collection) amortizes over an $8B+ book. The financial outcome that would deteriorate without this edge is the 11.9%–24.1% collection-minus-advance spread and the ~27% portfolio yield at acceptable loss rates. (INTERPRETATION; 10-K FY2025.)

What is not the moat. The dealer-holdback subordination is real incentive alignment (it makes the dealer co-underwrite), and it is patented and old — but it is a contract structure, not a barrier to entry. It disciplines adverse selection; it does not stop a competitor from offering the same or a simpler deal. Treat it as a business-model feature, not a moat.

Switching costs are LOW. The 10-K states plainly that a dealer “can cease assigning Consumer Loans to us at any time without terminating the Dealer servicing agreement.” Dealers are multi-homed — they shop each application across CACC, Westlake, Exeter, Santander, and captives via aggregators (RouteOne/Dealertrack). The 15% termination fee applies only to formal contract termination, not to simply routing volume elsewhere — which is exactly what is happening. (FACT — 10-K FY2025, Item 1.)

Network effects — weak, not self-reinforcing. The “more dealers → more data → better pricing → more dealers” loop exists but has diminishing returns once a deep statistical base exists, and rivals now have their own large subprime datasets plus third-party AI scoring. Label the network-effect claim speculative.

Market-share-stability test — FAILS (the single most damaging fact). Subprime market share fell from ~6.5% to ~5.1% over 2024→2025 (a >1.4-point drop), worsening to ~5.4%→4.5% in the core used segment in early Q4’25. A genuine moat should produce stable share; CACC is losing >1 point/year. The filing corroborates: FY2025 Consumer Loan unit volume fell 12.6% and dollar volume 16.5% even as active dealers rose 1.8%, because average volume per active dealer fell 14.4% (25.0→21.4) and same-store volume/dealer fell 14.3%. Dealers are staying enrolled but routing fewer deals to CACC — direct evidence of weak switching costs and lost competitiveness. (FACT — 10-K FY2025 MD&A volume tables; share figure is a third-party estimate.)

Cause — discipline vs. displacement (the open question). Two explanations coexist: (a) a deliberate Q3-2024 scorecard change that tightened underwriting and lowered advance rates (self-inflicted, defensible discipline — the 10-K notes “increasing advance rates… reduces the return on capital”), and (b) genuine competitive displacement by better-capitalized rivals chasing volume with abundant ABS funding. How much is voluntary discipline versus permanent share loss is the thesis question, and one quarter of recovery does not resolve it.

ROIC test — PASSES, so far. ROE was ~25.9% in FY2025 (filing-derived; ~28% on trailing third-party reads) on a 30±year record of high returns and per-share compounding; the forecasted collection-minus-advance spread ranged 11.9%–24.1% across the last 10 cohorts, with the 2025 Dealer-Loan spread at 22.5% vs 20.0% (2024). Returns remain well above cost of capital — advantages are present — but the failing share test says they are eroding.

Capital-cycle lens. CACC’s falling share and falling volume-per-dealer is exactly what supply-side capital flooding looks like: CACC is the disciplined incumbent pulling back on price while rivals chase volume on abundant ABS funding. That is capital-cycle-rational, but it costs share now, and the payoff (rivals over-advance → take losses → retreat → CACC re-takes share) is unproven this cycle.

Competitor comparison. CACC is differentiated by the Portfolio-Program risk-share/holdback model and its deepest-tier focus; most rivals buy loans outright (like CACC’s Purchase Program) and compete on advance rate, speed, and aggregator integration. Santander Consumer (the largest deep-subprime player) was taken private by its parent in 2022 at only ~1.0–1.1× book / ~6–7× earnings — a sobering reminder that the largest deep-subprime auto franchise was valued near book, not at CACC’s ~3.5–4×.

Competitive-position verdict — NARROW / ERODING advantage, not durable as currently demonstrated. The real moat (long-cycle proprietary loss data + scaled servicing) is genuine, but the dealer-holdback “moat” is a contract feature, switching costs are near-zero, and falling share is hard evidence the advantage is being competed away. A moat that cannot keep share is failing its central test. CACC is a good high-ROIC franchise in a structurally hard, capital-attracting industry under live capital-cycle pressure — a genuinely good business, but not a fortress. Durability is contingent on the underwriting-data edge proving un-replicable and management’s price discipline being rewarded by competitor retreat — neither yet evident in the data.

5. Growth History and Forward Opportunities

Historical growth — high-quality compounding, now stalling on volume. CACC’s history is one of disciplined organic growth and aggressive per-share compounding rather than balance-sheet empire-building. Revenue grew from ~$1.86B (FY2021) to $2.32B (FY2025), a ~5.7% CAGR, driven by a growing and re-priced loan book, not acquisitions. But the forward volume signal has turned negative: FY2025 Consumer Loan unit volume fell 12.6% and dollar volume 16.5% — the clearest sign that the growth algorithm (more dealers × more loans/dealer × healthy spread) has broken on the middle term. Active dealers keep rising (a record 15,745), but volume per dealer is falling faster — growth is being given back through the channel. (FACT — 10-K FY2025.)

Organic vs. acquired — 100% organic. No material M&A in the trailing three-year filing record or in company history. Growth has always been dealer enrollment + loan origination + book re-pricing. This is a positive (no diworsification, no integration risk) but also means there is no inorganic lever to offset organic volume softness.

Forward opportunities.

• RouteOne/DMS integration + the Hegde platform pivot. The most concrete forward initiative is integrating with dealer-management systems (RouteOne) to win franchise and larger dealers, attacking the share loss that the legacy standalone “CAPS” software is blamed for. New CEO Vinayak Hegde (ex-Airbnb/Amazon/T-Mobile, effective 2025-11-13) is explicitly a platform/UX hire, signaling management sees the problem as distribution and user experience, not credit. This is a credible diagnosis but an unproven strategy executed by a leader new to subprime credit. (INTERPRETATION; to be verified against the next 10-Q.)
• Survivor consolidation. If the Tricolor-led cleanse durably removes marginal capacity, CACC can re-take share at firm spreads — the late-bust payoff. Contingent and unproven.
• 2025 vintage outperformance. The 2025 vintage is underwritten at/above target (spread 22.2% vs 22.0%), so simply holding current underwriting quality while volume normalizes would restore earnings growth without needing share gains.

Growth verdict — historically high-quality, currently low-quality. The 30-year per-share record is real and high-quality (organic, high-ROIC, disciplined). But current growth is low-quality-to-negative: volume is shrinking, and the per-share growth now comes disproportionately from buybacks (denominator shrink) rather than franchise expansion. The forward opportunities are credible but speculative; the base case is stabilization, not re-acceleration.

6. Financial Quality

Multi-year spine ($M; FYE Dec 31; reconciled to SEC EDGAR XBRL + 10-K).

(FY2021 NI $958.3M is a stimulus / used-car-price anomaly — not a baseline.)

The collection-forecast engine — the most important analysis. CACC recognizes finance-charge income on a level-yield basis off forecasted collection rates; revisions to those forecasts flow through the P&L as a non-cash “provision for credit losses on forecast changes.” The vintage table is the whole story:

Three things this proves: (1) the 2022 vintage is the disaster — forecast collections collapsed from 67.5% to 59.3%, crushing the spread from ~20% to 11.9% (that book earns roughly half its intended margin); 2023 (−4.2) and 2021 (−2.5) are also impaired. (2) CACC over-advanced into a deteriorating consumer — the advance rate crept up to 46–47% on the 2021–2023 vintages (vs ~44% in 2016–2020) exactly as collection rates fell, which is why the spread compressed. (3) The model self-corrected — the 2024 (−1.9) and especially 2025 (+0.2, spread 22.2% > 22.0% target, advance back to 45%) vintages are clean. (FACT — 10-K FY2025 economics table.)

GAAP vs. economic earnings — partly fair, partly self-serving. The forecast-change provision peaked at $493.8M (FY2024) and fell to $338.3M (FY2025), but FY2025 still carried a $169.5M (1.5%) net-cash-flow forecast decrease — the drag is shrinking, not gone. Management’s “Adjusted/economic profit” framing is legitimate for the day-1/new-assignment timing distortion (CECL front-loads the provision on new loans, optically depressing GAAP NI when origination grows), but self-serving if used to wave away the forecast-change line, which is a real loss on money already advanced to the 2021–2023 books — permanent impairment, not a timing artifact. (INTERPRETATION.)

Latest quarter confirms recovery. Q1-2026 (vs Q1-2025): revenue $580.0M vs $571.1M; net income $135.8M vs $106.3M (+28%); the forecast-change provision fell to $54.4M (from $76.3M) — the lowest in years and the single best piece of stabilization evidence — though one quarter is not proof. (FACT — 10-Q filed 2026-05-05.)

ROE decomposition — real but flattered. FY2025 ROE ~25.9% ≈ ROA ~4.9% × equity multiplier ~5.3×. The ~5× leverage does the heavy lifting; the underlying ROA (~4.9%) is solid for a lender but not extraordinary. CACC also manufactures per-share ROE by shrinking the denominator — equity is held low because 100%+ of free cash flow goes to buybacks (equity fell from $1,749.6M to $1,523.6M in FY2025), so even flat NI lifts ROE. The ~26% is real but flattered by (a) recovering off a depressed 2024 base and (b) a shrinking equity base — not a clean steady-state read. (INTERPRETATION.)

Cash flow — why CFO >> GAAP NI. CFO was $1,054.6M in FY2025 (2.5× NI) and $1,137.9M in FY2024 (4.6× NI). The gap is the non-cash provision: GAAP NI is hit by forward-looking provisions ($616.1M FY2025) while collections keep arriving in cash regardless of the forecast mark. But CFO overstates distributable economics in the other direction — it ignores the cash that must be re-advanced to maintain the book; true owner cash sits between GAAP NI and CFO. Neither extreme should anchor valuation. (INTERPRETATION.)

Balance sheet, leverage, and funding. Debt ~$6,388M (secured/ABS $5,158.8M + senior notes $1,087.8M + revolver $107.3M); debt/equity ~4.19×. Cash collapsed to $22.8M (from $343.7M) — CACC runs minimal cash and funds via committed warehouse/ABS, with liquidity in undrawn facility capacity. The leverage is term-matched, largely non-recourse securitization against a self-liquidating book — so the real risk is funding cost, not a refinancing cliff: interest expense rose +178% (from $166.6M in FY2022 to $462.9M in FY2025) as cheap pre-2022 ABS rolled into higher-SOFR paper. That ramp is now inflecting (2025 warehouse cuts to SOFR+185, Nov-2025 ABS ~5.1%), which should decelerate or partly reverse the headwind. (FACT — 10-K FY2025; 8-Ks.)

Allowance. Net loans $7,909.2M; allowance $3,602.3M; gross ~$11.51B → implied lifetime allowance ~31.3% of gross. This high ratio is the deep-subprime model (CACC forecasts ~64–68% collection of contractual principal+interest), not distress per se. (FACT — 10-K FY2025.)

Unit economics (extracted from the FY2025 disclosures). The per-loan math makes the model concrete. FY2025 originations were 337,411 units against cash deployed of $3,856.1M (Advances to Dealers $2,764.8M + Purchases of Consumer Loans $1,091.3M), an average of ~$11,430 of cash put to work per loan at an advance rate of ~45% of the contractual amount — implying an average contractual receivable of ~$25,000, in line with the ~$27,500 industry average used-loan size. On that ~$11,430 advance, CACC forecasts collecting ~67% of the ~$25,000 contractual balance (~$16,750) over the loan’s ~5–6 year life, of which ~20% (~$3,350) is taken as a servicing fee before the advance is repaid — so the gross forecast spread is the ~22% of contractual balance the vintage table shows, or roughly $5,000+ of forecast collections above the advance per loan, against which CACC must absorb funding cost (now ~5% all-in on the marginal ABS), servicing/collection cost, and any forecast miss. The 2022 vintage shows what a miss does: an 8.2-point collection shortfall on a ~$25,000 contract is ~$2,000 of lost collections per loan — enough to roughly halve the spread. This is why forecasting accuracy, not volume, is the swing variable: the entire residual rent fits inside an ~8-point band of collection-rate error. (INTERPRETATION, derived from 10-K FY2025 deployment and vintage tables; per-loan figures are averages, not disclosed line items.)

SBC and dilution. SBC rose to $50.7M in FY2025 (12.0% of NI; 18.2% on the depressed FY2024 base). It is a real expense, not to be added back — but it is not the dilution story: the share count fell from 12.757M (FY2022) to 10.680M (FY2025), −16.3% in three years, as buybacks (1.514M shares in FY2025) vastly outweighed grant dilution. (FACT — SEC EDGAR XBRL; 10-K FY2025.)

Financial-quality verdict — high-quality cash economics, but returns are flattered and the spine is whipsawed by the forecast mark. Economics do not obviously improve with scale right now — the ROA is ordinary, the ROE is leverage- and buyback-amplified, and management’s own Economic Profit (see Capital Allocation) says true returns are compressing. The cash generation is genuine; the quality of the headline ROE is the caveat.

7. Capital Allocation

The machine. CACC has never paid a dividend; 100% of capital return is buybacks. Shares outstanding fell from 27.304M (2010) to 14.146M (2021) to 10.680M (2025) — −61% over 15 years, −24.5% over FY2022→FY2025. A textbook owner-operator float-shrink machine. The cadence shows genuine opportunism: heavy in 2021–22, paused in 2023 ($202.6M, the trough, when cheap funding disappeared), re-accelerated in 2024–25 ($313.3M → $725.4M). (FACT — SEC EDGAR XBRL PaymentsForRepurchaseOfCommonStock; 10-K FY2025.)

Buyback price vs. book — the central judgment. FY2025 repurchased 1,514,000 shares for $725.4M (~$479/share gross), at ~3.3–3.4× the year-end BVPS of $142.7. CACC has bought back at a premium to book for its entire modern history. This is the polar opposite of buying near/below book (mechanically accretive to per-share book). Buying at >3× book destroys per-share book value on the transaction — each retired share costs ~3.3× what it adds back in equity. The justification can only be intrinsic value, not book: at a normalized ~$40–50 EPS, ~$479 ≈ 10–12× earnings on a ~20%+ ROE book — defensible as value-accretive if the spread/ROE hold. The buyback is EPS-accretive but book-per-share-dilutive; intrinsic-value accretion rests entirely on the durability of the underwriting moat. (INTERPRETATION.)

Debt-funded re-leveraging. In FY2025, buybacks ($725.4M) were ~1.7× GAAP earnings and exceeded NI + SBC, so equity fell $226M while the loan book grew on borrowed money — a material slice of the buyback was effectively debt-funded re-leveraging. For a cyclical subprime lender in a soft-credit window, buying back stock at 3.3× book with borrowed money is the late-cycle behavior the capital-cycle lens flags as value-at-risk: it maximizes per-share upside if collections hold and magnifies downside if the 2022–2024 vintages deteriorate. The +178% interest-cost ramp is the visible cost. (INTERPRETATION.)

M&A — none. Purely organic; capital allocation = loan-book growth + buybacks, full stop. No empire-building, no diworsification — a clear positive.

Management incentives — best-in-class, and honestly signaling compression. The company-selected performance measure is Economic Profit (an EVA-style metric: Adjusted NI + after-tax adjusted interest, less a cost-of-capital charge on average capital). The 5-year reconciliation: $574.1M (2021) → $476.6M (2022) → $260.5M (2023) → $206.8M (2024 restated) → $173.3M (2025). This is a genuinely high-quality, rare structure — it charges management for the capital they consume and is keyed to long-run loan economics, not GAAP optics or revenue growth (the anti-“growth-for-growth’s-sake” design). Two caveats: (1) EP fell every year even as GAAP NI recovered in 2025 — the rising cost-of-capital charge (funding +178%) is eating the spread, so the comp metric is correctly signaling that true economic return is compressing; (2) Adjusted NI is built on management’s own collection forecasts, so the 2021–2023 over-forecasting means EP was likely overstated when comp was paid on it. CEO equity is granted upfront “in lieu of annual cash incentives” over a 10-year horizon — concentrating alignment but front-loading reward; new CEO Hegde’s 2025 package was $62.6M (incl. a $61.6M upfront option grant), a rich bogey. (FACT — 2026 DEF 14A.)

Insider ownership and behavior. Concentrated and structurally aligned: Prescott General Partners 21.2%, the Foss-Watson family 16.2% + ~7.6% in Foss trusts, directors & officers as a group 633,590 shares; public market-data aggregators report ~56% insider ownership. The named blocks reconcile to ~51%; a round “~65.8%” figure is an inclusive approximation that does not cleanly sum. Behavior is the caveat: a review of the trailing-three-year insider-filing corpus found zero code-P open-market purchases across 131 Form 4s, while the dominant real-money flow is large-holder selling (the late founder Don Foss’s estate trust unwinding a concentrated position; Prescott trimming; Jill Foss Watson). This is textbook diversification/non-information selling — not a red flag, but there is no bullish insider tell either. Alignment is structural (existing stakes + EP comp), not a fresh-conviction signal. (FACT — Form 4 corpus; 2026 DEF 14A. 10b5-1 status of the sales is unconfirmed.)

Capital-allocation verdict — disciplined, coherent, internally consistent, but a high-stakes PRO-CYCLICAL bet, not a low-risk compounder. The strengths are real (float-shrink, zero diworsification, opportunistic cadence, a best-in-class EP incentive that is honestly signaling compression). The risk is equally real: buybacks at ~3.3× book, ~1.7× earnings, partly debt-funded, into a deliberately shrinking equity base at ~4.2× leverage — management is re-leveraging a cyclical subprime lender to shrink the float into a softening credit window. Value-accretive on EPS and (if spread/ROE hold) on intrinsic value, but value-destructive on book-per-share and downside-amplifying. Not weak capital allocation; rather, concentrated risk dressed as discipline.

8. Changes and Headwinds — Last Two Years

Strategic / leadership. The most consequential change is the CEO transition: Kenneth Booth retired and Vinayak Hegde (a sitting director since 2021, with an ex-Airbnb/Amazon/T-Mobile platform/marketing background) became CEO effective 2025-11-13. Concurrently, the Chief Analytics Officer (Arthur Smith) and Chief Sales Officer (Daniel Ulatowski) retired effective 2026-02-01. A platform/UX leader inheriting a quant-underwriting business just as its long-time analytics chief departs is a genuine key-person / model-continuity risk into the most important part of the franchise — and it lands precisely during the 2024/2025-vintage-repair thesis that depends on the forecasting model. (FACT — 8-Ks 2025-10-25, 2026-01-20.)

Credit cycle. The 2021–2023 vintages over-forecast and were repeatedly marked down (forecast-change provisions of $413.7M / $493.8M in 2023–24), driving GAAP NI to a $247.9M trough in 2024; FY2025 ($423.9M) and Q1-2026 ($135.8M, +28%) mark the recovery, with the forecast-change provision shrinking. The swing factor is whether the 2022–2023 books have stopped slipping (Q1-2026’s $54.4M forecast-change provision says tentatively yes — but one quarter).

Funding. The +178% FY2022–25 interest-cost ramp is the dominant multi-year headwind; it began inflecting in 2025 (warehouse spreads cut to SOFR+185; Nov-2025 ABS ~5.1% to repay higher-cost debt). Do not extrapolate the historical interest-cost growth.

Regulatory / litigation. Materially eased: the CFPB withdrew as plaintiff (2025-04-29); Kansas, Texas, and Iowa withdrew from the multi-state investigation; CACC offered $45.0M (Sept-2025) and reached “preliminary alignment” on a ~$75.5M cash settlement (Jan-2026); cumulative contingent loss recognized is $82.6M — immaterial to a ~$424M-NI, ~$8.6B-asset company. The residual is the NY AG matter and a surviving nationwide putative class action (partial-dismissal ruling 2026-02-04). The NY FAIR Act (Dec-2025) is the new structural overhang. (FACT — 10-K FY2025 Legal Proceedings.)

Competitive shock. The Tricolor-led subprime cleanse (Ch. 7 + fraud charges) plus record subprime delinquencies define the live competitive backdrop — a survivor tailwind for CACC layered on a stressed consumer.

News / sentiment check. The curated recent-news tape on CACC is quiet — no thesis-relevant scored coverage as of 2026-06-05. A quiet tape is a valid, common finding; it is not thesis-changing here. The live signal that matters is in the filings and the credit data, not the headline flow.

Changes verdict — mixed, net modestly thesis-strengthening on the cyclical axis, thesis-weakening on the structural axis. Funding relief + regulatory de-risking + the recovery in earnings strengthen the near-term case; the leadership/analytics turnover and the continued share loss weaken the structural case. The changes do not resolve the central tension — they sharpen it.

9. Risk Analysis (Risk Matrix)

Risk of a catastrophic loss is low but non-zero: the funding is term-matched and largely non-recourse against a self-liquidating book, so there is no classic run/refi-cliff failure mode. The realistic severe-downside scenario is not bankruptcy but a multi-year compression of ROE and a P/B de-rate (R1+R2+R3+R4 compounding), which would impair the equity value materially without threatening solvency.

10. Valuation Discussion (Embedded Expectations)

Metric choice. For a deep-subprime levered balance-sheet lender, the correct lenses are: (1) P/E — trailing and normalized, since GAAP NI is whipsawed by the non-cash forecast-change provision; (2) P/B — the cleanest cross-cycle lender anchor, paired with (3) a ROE↔P/B justified-multiple bridge [P/B = (ROE − g)/(COE − g)]; and (4) Economic Profit (management’s EVA metric) as the through-the-noise read. EV/EBITDA is N/M and excluded — for a lender, interest is a cost of goods (funding), not a capital-structure item. Price-to-sales is low-information for a lender. (INTERPRETATION.)

Where CACC trades (2026-06-04/05, unofficial). ~$544–548; trailing P/E ~13.6–13.8×; forward P/E ~10.0–10.8×; P/B ~3.5× (on period-end equity) to ~4.0× (on the buyback-shrunk intra-year equity); P/S ~2.7×; no dividend. TTM EPS ~$40.24. 52-week range $401.90–$579.80 — the stock sits near the upper end after a ~25–30% re-rate from the ~$420–430 level of seven months ago.

Own-history valuation context. Despite the headline-premium P/B, CACC screens cheap versus its own history: composite valuation percentile ~37 (cheaper than ~63% of its own past), P/E percentile ~46 (mid), P/B percentile ~59 (slightly above its own median), and P/S percentile ~7.5 (near the cheapest it has ever been on sales), on a reliable 3-component reading. Compare a ticker only against its own past, never cross-sectionally — but on that basis the stock is not expensive relative to where it has historically traded. (Public market-data aggregators, 2026-06-04; reconcile to filings.)

Comp table (public market-data aggregators, 2026-06-05, unofficial — reconcile to filings; ROE approximate trailing).

Where CACC screens. It carries the highest P/B in the group by a wide margin (~3.5× vs next-highest WRLD ~3.0×; ALLY/COF/RM ~1×), yet only a mid-pack trailing P/E and a competitive forward P/E. The sharpest comp is SYF: a ~20–23% ROE consumer lender at only ~1.8× book — direct evidence the market does not mechanically pay 3.5× book for a low-20s% ROE. CACC’s premium is therefore not explained by ROE level alone; it pays for the 30-year per-share compounding record, the float-shrink machine, and historically higher (now-normalizing) returns.

The P/B-vs-P/E puzzle, resolved. A static Gordon-growth justified P/B [P/B = (ROE − g)/(COE − g), g = 2%] inverts to show what ROE the current ~3.5× book requires: at COE 10% ⇒ ~30% sustainable ROE; COE 11% ⇒ ~33.5%; COE 12% ⇒ ~37%. That is above the normalized ~22–26% ROE (itself flattered by leverage + a shrinking denominator) and far above the FY2024 ~14% trough. So on a clean steady-state basis CACC looks rich to book. The premium is reconciled — but not fully justified — by four things the simple model omits: (1) BVPS is artificially depressed because CACC retires stock at ~3.3× book, mechanically inflating both P/B and ROE; (2) the low-teens P/E says the earnings lens does not see a rich stock; (3) high-ROE reinvestment at ~0 payout compounds intrinsic value faster than g=2% assumes; (4) a track-record/scarcity premium. Net: the P/E (~10–13.6×) is the load-bearing anchor; the 3.5× book is a derived artifact of the buyback policy, not independent evidence of cheapness or a downside floor. If ROE settles at ~20–24% (not 30%+), the static-model fair P/B is ~2.0–2.9× — i.e., book offers less downside cushion than the headline premium implies, and the thesis rests on earnings durability, not asset value.

Embedded expectations — what ~$544 requires the market to underwrite jointly:

• (a) Spread/collection stability: the clean 2024/2025 vintages (spread ~22%, variance ~0/+0.2) are the true run-rate and the 2021–2023 over-advance episode does not recur. (Q1-2026 forecast-change provision $54.4M corroborates.)
• (b) Normalized NI ~$450–550M sustained ≈ ~$42–52 EPS on the current share count, putting ~$544 at ~10.5–13× normalized earnings — a recovery multiple, not a deep-value one; the re-rate has banked much of the recovery.
• © Funding-cost relief (2025 SOFR+185 inflection) reverses enough of the +178% interest ramp to stop net-spread compression. (But Economic Profit fell to $173M — the market is betting that EP trough is in.)
• (d) Buyback continuation (~12% float shrink/yr at ~3.3× book) so even flat NI lifts EPS — the engine that makes the premium-to-book defensible without a 30% ROE, but it is debt-funded re-leveraging that amplifies downside.

What the market is pricing correctly: the funding inflection (documented in 8-Ks); the clean 2024/2025 vintages (in the filing); the regulatory easing; and the real, EPS-accretive buyback. What it may be pricing incorrectly: treating ~26% ROE as the run-rate when EP says returns are compressing; extrapolating the vintage self-correction through a full cycle while the consumer is at record delinquency and share is still falling; and assuming ~3.5× book is a floor when a 20–24% ROE justifies only ~2.0–2.9×.

Scenario analysis (embedded-expectations frame — NOT a price target).

Valuation verdict. The ~3.5× P/B is only “correct” in the bull case (28–32% ROE). In the base case the justified P/B is ~2.3–2.9×, so at today’s ~3.5× the market is pricing closer to bull-than-base on the asset lens, while the earnings lens (~10–13× normalized) is merely fair. The asymmetry is unfavorable in the bear case because the de-rate hits both the multiple and the earnings. The buyback is the double-edged embedded assumption: it manufactures the per-share compounding the bull needs, but it is debt-funded re-leveraging that magnifies the bear.

11. Variant Perception

What consensus believes. Roughly: “the crisis has passed.” The CFPB is gone, credit has bottomed, the 2025 vintage is clean, funding cost is inflecting down, and a 30-year owner-operator buyback machine is compounding per-share value — so the FY2024 trough was the bottom and ~$544 is reasonable for a recovering high-ROE franchise. The ~$537 third-party analyst target and a “hold”-ish ~2.5 analyst rating sit right at the current price — consensus sees the stock as roughly fairly valued, not a screaming opportunity. (Analyst figures are third-party market color, not a price target.)

The strongest bull case. This is a survivor in a self-cleansing bad neighborhood: the deep-subprime tail is losing its weakest (often fraudulent) capacity (Tricolor, ACC, Automotive Credit Corp) while demand persists and collateral values are firm. CACC gets cheaper funding as rivals get cut off, the federal regulator has retreated, and the 2025 vintage proves the underwriting model self-corrected. A platform CEO can re-accelerate volume via RouteOne, and the buyback retires ~12% of the float a year at what — on normalized earnings — is ~10× P/E. The stock is cheap versus its own 10-year history (composite ~37th percentile), and ~29% of the float is short — an owner-operator buying back this aggressively can squeeze the bears.

The strongest bear case. This is a structural decline dressed up as a cyclical recovery. Market share is still falling (6.5%→5.1%, volume/dealer −14.4%), AI underwriting and aggregator distribution are permanently commoditizing CACC’s data edge, and the “moat” fails the share-stability test. The ~26% ROE is leverage- and buyback-flattered; management’s own Economic Profit fell every year to $173M, signaling true returns are compressing. The premium ~3.5× book requires a 30%+ perpetual ROE the business no longer earns, and buybacks at 3.3× book partly on debt are re-leveraging a cyclical lender into a record-delinquency consumer environment — late-cycle behavior. SYF earns the same ROE at half the multiple.

The crowded short (~29% of float). This is itself the variant perception in sharpest relief: a large cohort is betting on the bear case (structural decline + over-levered buyback into a credit downturn). The configuration cuts both ways — it confirms the bear thesis is mainstream and sets up squeeze risk against a relentless corporate buyer of its own stock. (FACT — short interest ~28.75% of float, public market-data aggregator 2026-06-05.)

The 3–5 assumptions that matter most: (1) Are the 2024/2025 vintage spreads (~22%) the durable run-rate, or a head-fake before another slip? (2) Is the share loss voluntary discipline that mean-reverts, or permanent competitive displacement? (3) Does funding cost keep easing, or was 2025 a brief window? (4) Is the sustainable ROE ~20–24% (justifies ~2–2.9× book) or ~28–32% (justifies ~3.5×)? (5) Does the platform-CEO transition, amid analytics-team turnover, preserve or degrade the forecasting model?

What would falsify each side. Bull falsified by: the forecast-change provision re-accelerating over the next two 10-Qs (2024/2025 vintages slipping) and/or share dropping below ~4.5%. Bear falsified by: two more clean quarters on the 2024/2025 vintages plus a stabilization or recovery in volume-per-dealer — proof the model self-corrected and the franchise stopped shrinking.

12. Fact vs. Interpretation Table

13. Open Questions

1. 10b5-1 status of the Foss-trust / Prescott / Watson code-S sales (Form 4 footnotes flag no plan; check Form 144 cover pages). If discretionary, it modestly weakens the “aligned owners” narrative.
2. Have the 2022–2023 vintages truly stopped slipping? One quarter (Q1-2026 forecast-change provision $54.4M) is supportive but not proof — watch the next two 10-Qs.
3. Is the share loss voluntary discipline or permanent displacement? The filing cannot separate the two; the volume-per-dealer trend over the next year is the tell.
4. Sustainable ROE and the right COE for a 4.2×-levered, politically targeted deep-subprime lender — the whole “is the premium justified” answer is COE-sensitive (10% vs 12–13% flips the justified P/B materially).
5. ABS maturity ladder / undrawn facility headroom — quantify the funding runway and the go-forward incremental funding spread (the 8-Ks show the inflection but not the full ladder).
6. Will the platform-CEO + analytics-team transition preserve the forecasting model’s accuracy, or introduce model drift at the worst possible time?
7. How far does the state-AG migration (NY FAIR Act) extend the regulatory tail beyond the ~$75.5M NY settlement?

14. What Must Be True (Bull and Bear, with Falsification Tests)

For the BULL case to be right, all of these must hold:

• The 2024/2025 vintages sustain ~22% spreads through a full credit cycle (the model self-corrected durably).
• Market share stabilizes and ideally re-accelerates as marginal competitors exit (the survivor payoff materializes) and the RouteOne/Hegde pivot wins back volume-per-dealer.
• Funding cost keeps easing (the 2025 inflection is the start of a trend, not a window), stabilizing the net spread and arresting the Economic-Profit decline.
• The buyback keeps compounding per-share value at a price that proves intrinsically accretive (i.e., ~26%+ ROE is real, not a leverage/denominator artifact).
• Falsification test: the forecast-change provision re-accelerates over the next two 10-Qs, or share falls below ~4.5%, or Economic Profit makes a new low — any one breaks the bull.

For the BEAR case to be right, all of these must hold:

• The 2024/2025 vintages start slipping like 2022 (spread → mid-teens), re-igniting forecast-change provisions and re-collapsing GAAP NI toward the ~$250–320M zone.
• Share loss continues (AI underwriting + aggregators permanently commoditize the data edge) and volume-per-dealer keeps falling.
• The premium ~3.5× book de-rates toward the ~1.5–2.0× a ~14–18% ROE justifies, hitting the multiple and the earnings simultaneously.
• Debt-funded buybacks at 4.2×+ leverage amplify the downturn rather than cushion it.
• Falsification test: two more clean vintage quarters plus a stabilization/recovery in volume-per-dealer would break the bear (the franchise stopped shrinking and the model held).

15. Source Appendix

See Appendix B below for the full, authority-ranked citation set — primary SEC filings, public data reconciled to filings, and secondary industry/regulatory sources.

Appendix A — Diligence Questionnaire

Company: Credit Acceptance Corporation (NASDAQ: CACC) | Date: 2026-06-05 Scope: Standing diligence questionnaire. Fact/Interpretation/Assumption labels used where it matters. No BUY/SELL or price target appears here.

General

What thoughtful questions have other investors asked about this company? The recurring, sophisticated questions are: (1) Is CACC’s GAAP earnings number meaningful at all, given finance charges are recognized on a level-yield basis off management’s own collection forecasts, with revisions hitting the P&L as non-cash provisions? (2) Is the dealer-holdback/Portfolio-Program structure a genuine moat or just a clever contract that anyone can copy? (3) Is the 30-year ROE record durable, or is it leverage + buybacks compounding a regulatory-arbitrage spread that is now mean-reverting? (4) Why does it trade at ~3.5× book when the largest peer (Santander Consumer) was taken private at ~1× book and SYF earns the same ROE at ~1.8×? (5) Is the buyback-funded float shrink intrinsic-value creation or financial engineering that masks a contracting franchise? The ~29%-of-float short interest shows how mainstream the skeptical view is.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: neither — recovering off a cyclical trough. FY2024 NI ($247.9M) was the trough (driven by 2021–2023 vintage forecast markdowns); FY2025 ($423.9M) and Q1-2026 (+28%) are the recovery. FY2021 ($958.3M) was a stimulus/used-car-price anomaly and is not a baseline. Normalized NI is judged ~$450–550M.

Are earnings driven primarily by the external environment or internal company actions? Both, inseparably. External: the deep-subprime consumer’s health, used-vehicle collateral values, funding-market spreads, and competitor advance-rate behavior. Internal: underwriting discipline (the scorecard/advance-rate decisions that caused the 2022 over-advance and the 2024/2025 self-correction), servicing/collections execution, and capital allocation. The forecast-change provision is the visible interface between the two.

How stable are revenues? Revenue is structurally recurring (finance charges recognized over the multi-year life of a self-liquidating book), so reported revenue is smoother than originations — FY2023→FY2025 revenue rose every year ($1,901.9M → $2,162.4M → $2,317.2M) even as FY2025 unit volume fell 12.6%, because the book and yield carried it. The quality of revenue (the spread) is far less stable than the top line.

Outlook for the company’s products and services? Persistent demand (subprime is ~22.5% of used-vehicle financing and rising), but CACC’s share of that demand is falling. The product (deep-subprime indirect financing) is not going away; CACC’s competitive claim on it is the question.

How big will this market be — growing, shrinking, domestic or international? Domestic-only. US auto debt is at a record ~$1.68T; the subprime-used pool is large and modestly growing. Market size is not the constraint — competitive share and credit quality are.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Interpretation: more competitive structurally (AI underwriting, aggregator distribution, banks/captives cycling down-market on abundant ABS funding), but temporarily less competitive at the deepest tail as fraudulent/marginal players (Tricolor, ACC, Automotive Credit Corp) exit. Net: a cyclical reprieve inside a secular intensification.

How profitable is this business (ROIC/ROE)? ROE ~25.9% FY2025 (filing-derived) ≈ ~4.9% ROA × ~5.3× leverage — high, but leverage- and buyback-flattered. The cleaner economic read (management’s Economic Profit, which charges for capital) is compressing: $574M (2021) → $173M (2025).

How profitable is the industry — competitors, barriers to entry? Fragmented, no dominant deep-tier firm; industry-level barriers are weak (commoditized capital, cyclical entry/exit). Any durable profit is firm-specific (loss-forecasting data + servicing scale + cheap funding).

Can the business be easily understood? The mechanics (advance vs. collect, spread, waterfall) are understandable; the earnings are not transparent because they depend on management’s forward collection forecasts. Medium-low transparency.

Can it be undermined by foreign, low-cost labor? No — domestic regulated lending and collections; labor arbitrage is not the threat. Technology (AI underwriting) is.

Do brands matter? Minimally to the consumer (the dealer is the customer); CACC’s “brand” with dealers is service level + a method to finance hard credits, not consumer brand equity.

Nature of competition / customers’ switching costs / barriers to entry. Competition is on advance rate, speed, service, and aggregator integration. Dealer switching costs are near-zero — a dealer can stop routing volume to CACC at any time without terminating the agreement, and dealers are multi-homed. Barriers to entry at the segment level are low; CACC’s edge is execution, not exclusion. (FACT — 10-K FY2025.)

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? Interpretation: the ~50-year proprietary loss dataset and servicing platform (the real moat) are intangibles not capitalized — economic value not on the balance sheet. Conversely, the loan book is carried net of a large (~31% of gross) allowance, so reported net loans are conservative if the collection forecasts are right.

Off-balance-sheet liabilities? The ABS securitizations are largely non-recourse and consolidated; the principal contingent off-book item is litigation (NY AG / class action; $82.6M cumulative contingent loss recognized, ~$75.5M preliminary cash settlement). No material operating-lease or pension surprises surfaced.

How conservative is the accounting? Mixed. The CECL/level-yield framework front-loads provisions (conservative on timing) but recognizes income off management’s own forecasts (aggressive in that it embeds judgment). The 2021–2023 over-forecasting — later clawed back — shows the judgment can be optimistic. Treat GAAP NI and management’s “Adjusted NI”/“Economic Profit” with equal skepticism; reconcile to cash collections.

How CapEx-hungry is the business? Physical CapEx is light (it is a finance/servicing company). The real “capital” hunger is the cash advanced to dealers / paid for loans ($3,856.1M deployed in FY2025) funded by securitization — the balance-sheet intensity is in the loan book and its funding, not PP&E.

Capital Allocation & Management

How much free cash flow, and how is it used? CFO was $1,054.6M (FY2025), well above GAAP NI, but a large share must be re-advanced to maintain the book — true distributable cash sits between NI and CFO. Discretionary capital return is 100% buybacks ($725.4M FY2025; never a dividend).

Philosophy? Owner-operator per-share compounding + tax efficiency for a concentrated family/affiliate base; opportunistic cadence (paused buybacks in the 2023 trough, re-accelerated 2024–25).

Significant acquisitions recently? None — purely organic, no material M&A ever. A positive (no diworsification).

Buying back shares? Aggressively — −61% shares since 2010, −24.5% over FY2022–25, 12.6% of the float in FY2025 alone — but at ~3.3× book and partly debt-funded, which is EPS-accretive and book-per-share-dilutive (intrinsic accretion hinges on ROE durability).

Issuing large amounts of new shares to insiders? SBC is real (~$50.7M FY2025, ~12% of NI) but dwarfed by buybacks; net share count is shrinking ~5–6%/yr. New CEO Hegde received a $61.6M upfront option grant (in lieu of ~10 years of annual grants) — concentrated, not recurring dilution.

Compensation policy of directors and management? Keyed to Economic Profit (EVA, charges for capital) — a rare, high-quality, anti-growth-for-growth’s-sake structure; the CEO is excluded from profit-sharing and paid via long-dated upfront options. Caveat: EP is built on management’s own forecasts.

Motivations of management? Structurally aligned with per-share intrinsic value via large existing stakes + EP comp. Caveat: alignment is structural, not a fresh-conviction signal — zero insider open-market buys in 36 months, and the dominant flow is large-holder diversification selling.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? No — a US C-corporation common stock (NASDAQ: CACC), standard 1099 treatment.

Dividend policy? None, ever. All capital return via buybacks.

How profitable is the business? ~26% ROE / ~4.9% ROA FY2025 — high but leverage/buyback-flattered and (per Economic Profit) compressing.

Is net income diverging from cash from operations? Yes, materially — CFO is 2.5–4.6× GAAP NI because the non-cash provision depresses NI while collections arrive in cash. This divergence is structural to the level-yield model, not a red flag by itself — but it cuts both ways (CFO overstates distributable cash because of required re-advances). The honest owner-earnings figure is between the two.

Other market data: ~$544–548; trailing P/E ~13.6×, forward ~10×, P/B ~3.5–4.0×, P/S ~2.7×; 52-week $401.90–$579.80; short interest ~29% of float; ~56% insider/affiliate ownership; cheap vs. its own 10-year valuation history (composite ~37th percentile). (Third-party analyst target ~$537 is market color, not a price target.)

Risks & Downside

What factors would cause the stock to decline? A re-acceleration of the forecast-change provision (2024/2025 vintages slipping like 2022); continued/accelerating share loss; funding-cost re-escalation; a worsening deep-subprime consumer; adverse state-AG/regulatory action; and a P/B de-rate toward the ~2.0–2.9× a normalized ROE justifies. Most of these compound.

Risk of a catastrophic loss? Low but non-zero. Term-matched, largely non-recourse securitization against a self-liquidating book means no classic run/refi-cliff failure mode. The realistic severe case is a multi-year ROE compression + P/B de-rate (equity-value impairment), not insolvency.

Chance of a total loss? Very low absent a severe, sustained multi-year credit and funding shock that overwhelms the non-recourse structure and the equity cushion — not the base or even bear case.

Recent News & Events

Has the business environment changed recently? Yes, on several axes (last ~18 months): (1) Leadership — CEO Booth → Hegde (platform/marketing background, eff. 2025-11-13) plus Chief Analytics + Chief Sales officer retirements (Feb-2026) — a key-person/model-continuity risk. (2) Regulatory — CFPB withdrew as plaintiff (Apr-2025); KS/TX/IA exited the multistate; ~$75.5M preliminary NY settlement; NY FAIR Act enacted (Dec-2025). (3) Credit — 2024/2025 vintages underwritten cleanly after the 2021–2023 slip; forecast-change provision shrinking. (4) Funding — warehouse spreads cut to SOFR+185, Nov-2025 ABS ~5.1% — the +178% interest ramp inflecting. (5) Competition — Tricolor-led subprime cleanse + record subprime delinquencies. The curated recent-news tape is otherwise quiet — no thesis-changing headline flow as of 2026-06-05.

Any significant acquisitions? None.

Recent change in accounting policies? No new policy change surfaced; the CECL/level-yield framework is unchanged — the inputs (collection forecasts) move, not the method.

Recent changes in the business — new markets, facilities, management? No new markets/geographies (US-only); the strategic change is the RouteOne/DMS platform integration to win franchise/larger dealers (the Hegde-era distribution pivot), aimed at the share loss blamed on the legacy CAPS software. Unproven.

Framework Lenses Applied

• Moat type (Greenwald taxonomy): economies of scale + customer captivity in deep-subprime loss-forecasting data and servicing; the dealer-holdback is a contract feature, not a barrier. Market-share-stability test: FAILS (share 6.5%→5.1%). ROIC test: passes (ROE >> COE) but eroding. EPV would lean on a normalized spread/ROE well below the buyback-flattered headline.
• Capital cycle (Marathon lens): the deep-subprime tail is in late-bust/early-recovery (marginal-capacity exit), but the broad auto-ABS window is at record volume — not a clean trough. Asset-growth discipline is good (CACC is pulling back on price), but the buyback-funded re-leveraging into a cyclical lender is the capital-cycle risk flag. Regulation distorts the normal cycle (federal retreat = near-term tailwind; state-AG migration = fatter tail).

Appendix B — Source Appendix

Company: Credit Acceptance Corporation (NASDAQ: CACC) | CIK 0000885550 | FYE Dec 31 | HQ Southfield, MI Date: 2026-06-05 Scope: Public primary and secondary sources supporting every load-bearing claim in this memo. Sources are ordered by authority: (1) primary SEC filings, (2) public data reconciled to filings, (3) secondary industry/regulatory. EDGAR base for all filings: https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0000885550&type={FORM}.

1. PRIMARY — SEC FILINGS

1.1 Annual reports — Form 10-K

EDGAR: https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0000885550&type=10-K

1.2 Quarterly reports — Form 10-Q (through Q1-2026)

EDGAR: .../type=10-Q

1.3 Proxy statements — DEF 14A / DEFA14A

EDGAR: .../type=DEF+14A

1.4 Material-event reports — Form 8-K (material items cited)

EDGAR: .../type=8-K

1.5 Securitization & other primary filings

EDGAR: .../type=4 (insider), .../type=ABS-15G

2. PRIMARY — DATA SOURCES (reconciled to filings)

3. SECONDARY — INDUSTRY / REGULATORY

All accessed 2026-06-05 unless noted. Where these summarize a primary document (e.g., Experian release, court docket), the primary governs.

3.1 Market sizing & collateral

• NY Fed Consumer Credit Panel / Equifax — auto loan+lease balances ~$1.68T (Q1 2026), Q4-2025 originations ~$180.8B. (Primary = the NY Fed Household Debt & Credit report.)
• Experian, “State of the Automotive Finance Market,” Q4 2025 — total subprime 15.31% of vehicle financing; used-subprime 22.47%; used avg loan $27,528. experian.com/blogs/insights/; experianplc.com newsroom, Feb 2026.
• Cox Automotive / Manheim, “Used Vehicle Value Index: May 2026 Trends” — MUVVI 212.6, +3.6% YoY, +0.3% MoM. coxautoinc.com, May 2026.
• Federal Reserve FEDS Note, “Subprime Auto Lending: Trends in Buy Here Pay Here Auto Lending” — federalreserve.gov, 2026-05-08.

3.2 Credit performance / ABS market

• Fitch Ratings (via Auto Finance News, autofinancenews.net) — subprime 60+ day DQ record ~6.9% (Jan 2026; 32-yr / 385-month high, series to 1993).
• S&P Global Ratings, “US Auto Loan ABS Tracker” (Oct 2025) — spglobal.com; full-year 2025 subprime ABS losses ~6.18% (vs 5.78% 2024); subprime tranche spreads BB ~425–575 / B ~575–850 bps.
• S&P Global Ratings via Auto Remarketing (autoremarketing.com) + Asset Securitization Report (americanbanker.com/asreport) — record ~$127B auto ABS in 2025 (2nd consecutive record; 34% of all ABS).
• Wolf Street (wolfstreet.com), Q4-2025 / Q1-2026 delinquency posts, Feb–May 2026 — corroborating DQ data.

3.3 Competitive exits / fraud (the capital-cycle evidence)

• CNN Business, “A major subprime auto lender just went belly up” — cnn.com, 2025-09-12 (Tricolor Holdings Ch. 7, 2025-09-10).
• CNBC, “Tricolor execs charged with ‘systematic fraud’” — cnbc.com, 2025-12-17 (founder/CEO Daniel Chu + execs federally charged; ~$800M fictitious collateral). (Primary = DOJ charging documents.)
• Auto Finance News — Tricolor ripple-effect / ABS fallout; Automotive Credit Corp pause (2025-08-07); American Car Center Ch. 7 (Mar 2023). autofinancenews.net, 2025–26.
• CACC funding bifurcation: StockTitan (stocktitan.net, 2025-11-13, CACC $500M ABS ~5.1%) and AInvest (ainvest.com, Sept 2025, $200M warehouse → SOFR+185). These are secondary relays of CACC 8-Ks — the 8-Ks (section 1.4 above) are the primary source and govern.
• CACC subprime market share 6.5%→5.1% (2024→2025), core used ~5.4%→4.5% early Q4-25: Auto Finance News and Auto Remarketing “Top 20 finance companies market share Q4.” autofinancenews.net; autoremarketing.com. Secondary-only — not in any CACC filing; the filing-based corroboration is the FY2025 volume-per-dealer decline (−14.4%).
• ROE ~28.7%: Macrotrends / Simply Wall St, macrotrends.net/stocks/charts/CACC/credit-acceptance/roe. Third-party computed — the filing-derived FY2025 ROE (~25.9% NI/avg equity) is the authoritative figure.

3.4 Regulatory / legal

• CFPB withdrawal: American Banker (“As auto delinquencies rise, CFPB seeks to cut oversight”); federalregister.gov 2025-05-12 withdrawal notice; consumerfinance.gov. Primary = the court docket / 10-K Legal Proceedings (CFPB unopposed motion to withdraw filed 2025-04-24, granted 2025-04-29).
• CFPB supervision-threshold rule (10,000 → 1,000,000 loans/yr; 63 → ~5 supervised): federalregister.gov proposed rule (Aug 2025); Covington “Funding Haze… The CFPB in 2026” (cov.com, 2026-01); Brownstein “CFPB Drops Indirect Lending Case” (bhfs.com).
• NY FAIR Business Practices Act (signed 2025-12-19): Consumer Financial Services Law Monitor (consumerfinancialserviceslawmonitor.com, 2026-01); Morgan Lewis (morganlewis.com, 2026-01). NY AG remains sole CACC plaintiff. Primary = the enacted NY statute + 10-K Legal Proceedings.
• Settlement trajectory ($45.0M offer Sept-2025 → $75.5M preliminary alignment Jan-2026; $82.6M cumulative contingent loss; KS/TX/IA withdrew; class action partial ruling 2026-02-04): CACC FY2025 10-K Legal Proceedings + MD&A — primary.

No BUY/SELL recommendation or price target appears in this appendix. Primary sources govern over all secondary relays.