Chapter 7

China credit quality through the 2025 cycle

The bear pillar of the discount is deteriorating credit, and the headline confirms it: China 90+ day delinquency rose from 1.98% at end-2023 to 2.85% at end-2025 and 3.11% by March 2026. That measure lags. The leading indicators — day-1 delinquency, cohort loss, collection rate — worsened on the second-half-2025 regulatory shock, then turned in mid-December and improved into early 2026. Reported and forward signals now point opposite ways.

Why credit risk lands on FinVolution's income statement

For most loans it facilitates in China, FinVolution keeps the credit risk rather than passing it to the funding bank — the quality-assurance mechanism examined in Guarantee Economics. The consequence for this chapter is direct: loan losses are not a footnote, they are the largest cost the company carries. In 2025, credit losses for the quality-assurance commitment (¥3,462.4 million), the provision for loans receivable (¥637.7 million) and the provision for accounts receivable and contract assets (¥426.0 million) together came to roughly ¥4.5 billion of charges [1] — more than the ¥4,124.9 million of guarantee income booked the same year [2]. Where delinquency goes, near-term earnings follow. That is why the credit trajectory, more than volume or take rate, is the swing factor on the 2026 profit line.

The lagging measure keeps rising

FinVolution's most-quoted asset-quality number is the 90+ day delinquency rate: the balance of China loans it bears risk on that are 90 to 179 days past due, as a share of the risk-bearing book. Loans 180 days or more overdue are treated as charged off and drop out of the ratio [3]. It has risen every year since 2023 and kept climbing into 2026.

Source: 90+ rates for Dec-2023 to Dec-2025 from FY2025 Form 20-F, Loan Performance Data [4]; March-2026 print from the Q1 2026 results release [5].

Two mechanics make this stock measure a poor read on what new lending is doing. First, it lags: a loan written today does not enter the 90+ bucket for at least three months, so the ratio reports underwriting decisions made two to three quarters earlier. Second, the denominator is shrinking. FinVolution has been pulling back its China risk-bearing book — the quality-assurance receivable fell from ¥1.6 billion at end-2024 to ¥1.3 billion at end-2025 [6] — so a roughly steady stock of seasoned bad loans becomes a larger percentage of a smaller base. Some of the rise in the ratio is arithmetic, not fresh deterioration. The number is real and it is going the wrong way, but it is the rear-view mirror.

The leading signal turned in December

The forward-looking indicators tell a cyclical story rather than a structural one. Day-1 delinquency — the share of principal due that goes unpaid on its first due date — and the 30-day collection rate move months ahead of the 90+ bucket. FinVolution plots both.

Source: quarterly series 1Q23–4Q25 from the March 2026 investor presentation [7]; 1Q26 (5.2%) from the Q1 2026 earnings call [8].

The pattern is a clean cycle. Day-1 delinquency bottomed around 4.7% in the first half of 2025, with the 30-day collection rate steady near 89%. Then China's mid-2025 consumer-lending regulation — the interest-rate-cap and liquidity tightening detailed in Overseas Engine — squeezed the industry. Management's expected vintage loss climbed from 2.5% to 2.9% in the third quarter as day-1 delinquency rose to 5.0% and the collection rate slipped to 88%, and take rate fell because credit costs rose [9]. It got worse into the fourth quarter: day-1 delinquency reached 5.5% and the collection rate fell to 86% [10].

The inflection came in December. Management reported the early risk indicators peaking in the middle of that month, with the newly originated cohort's vintage loss stabilizing around 3% [11], and day-1 delinquency trending down through January and February toward 5% [12]. By the first quarter of 2026 the improvement had carried through every metric: expected vintage loss eased from 3.0% to 2.7%, day-1 delinquency from 5.5% to 5.2%, the 30-day collection rate recovered from 85.9% to 86.8%, and the M2 flow-through rate fell from 0.77% to 0.68% [13]. Management said day-1 delinquency fell below 5% by April, back to where it sat before the shock [14].

Sources: day-1 delinquency and 30-day collection rate as plotted in the March 2026 investor presentation [15]; management's per-quarter expected vintage loss from the Q3 2025 [16], Q4 2025 [17] and Q1 2026 [18] calls.

The cohort curves say underwriting held

The strongest evidence that this is a cycle rather than a break is the vintage curve — the cumulative loss experience of each quarterly loan cohort as it ages, month by month. FinVolution's curves are strikingly tight: every cohort from the first quarter of 2022 through the third quarter of 2025 tracks to roughly 0.6% by month three, 1.9% by month six and 2.8% by month nine [19]. Measured at full 12-month maturity, the realized vintage delinquency rate was 2.84% for loans written in 2023 and 2.32% for those written in 2024 — an improvement, not a deterioration, in the quality of new lending across those two years [20].

The honest limit sits in the same table. The 2025 cohorts — the ones written into and through the regulatory shock — have only begun to season. FinVolution reports their blended vintage rate at 1.69% as of December 2025, but flags that this covers only their early months and will rise as the loans age toward the 12-month mark [21]. Management's own real-time estimate for those cohorts touched 3.0% before easing. So the underwriting record through 2024 is genuinely good, and the 2025 vintages are the open question — their final loss is not yet knowable from the curves.

Where the reserves are moving

Total credit charges actually fell in 2025, which is easy to misread. Credit losses for the quality-assurance commitment dropped 24.5%, from ¥4,587.3 million to ¥3,462.4 million — but the company is explicit that this reflects a smaller risk-bearing book in China, not lower losses per loan [22]. At the same time the provision for loans receivable nearly doubled, up 99.3% to ¥637.7 million, as FinVolution held more loans on its own balance sheet: loans receivable grew 55.7%, from ¥4.2 billion to ¥6.5 billion [23].

Source: FY2025 Form 20-F, Results of Operations [24]; year-on-year drivers on p.117 [25].

The signal is a shift in where the risk sits. As FinVolution shrinks the off-balance-sheet guarantee book and grows the on-balance-sheet portfolio, the on-balance-sheet provision line is where deterioration shows up first and most visibly. The first quarter of 2026 makes the point: the provision for loans receivable was ¥218.1 million, against ¥85.4 million a year earlier — up more than two and a half times — as the held loan balance rose [26]. Part of that jump is a bigger book; part is a higher loss rate on it. Group net profit fell to ¥421.1 million in the quarter, from ¥737.6 million, and credit charges were the largest reason [27].

A sector event, read against peers

The 2025 deterioration was not FinVolution-specific. Qifu, the scale leader in the peer set, saw its 90+ delinquency rate move the same way — 2.35% in 2023, down to 2.09% in 2024, then up to 2.71% in 2025 — and attributes the increase directly to the industry-wide liquidity squeeze that followed China's 2025 interest-rate-cap measures [28]. X Financial fared far worse: its 91-to-180-day delinquency rate jumped from 2.48% to 6.31% across the same window [29].

Sources: FinVolution 90+ rate from FY2025 Form 20-F [30]; Qifu from its FY2025 Form 20-F [31]; X Financial 91–180-day rate from its FY2025 Form 20-F [32].

FinVolution sits in the middle of that range: modestly worse than Qifu on the same headline number, far better than X Financial. The deterioration reads as a cyclical, policy-driven event that hit the whole China consumer-lending industry, and FinVolution absorbed it in line with the better-run half of its peers. The company-specific question is not whether it cracked — it did not — but whether the early-2026 recovery in its leading indicators sustains.

What the constructive read rests on

The evidence points to a credit cycle that peaked around December 2025 and has been improving since: the vintage curves show no break in underwriting through 2024, day-1 delinquency and collection rates have recovered two quarters running, and the deterioration was a shared industry event rather than a FinVolution failure. The strongest fact against that read is that the constructive case is only two quarters old, on management's own telling — the company itself called the November move "too early to draw a conclusion" [33] — while the 90+ ratio and the on-balance-sheet provision, the numbers that actually hit reported profit, are still rising because they lag.

Three credit-specific markers would decide it, each a line an investor can check in the next two filings: whether day-1 delinquency holds below the ~5% pre-shock level rather than re-accelerating; whether the 2025 cohorts season out near the 2.7% expected rate or drift toward the 3.0% that management briefly flagged; and whether the 90+ ratio, which mechanically has further to climb as the shock cohorts age, tops out by mid-2026 rather than pushing higher. On the answer to those turns much of whether the discount examined in Valuation is a floor under a stabilizing book or a warning about one still rolling over.