When South Africa's big five banks reported their 2025 annual results, the agricultural credit book looked, on the surface, unremarkable. Headline NPL ratios for agri portfolios sat between 3.8% and 5.1%. Impairment charges were elevated but not alarming. The story the numbers told was one of managed stress, not structural rupture.
That story is incomplete. Strip out the interest rate tailwind that has mechanically boosted interest income over the past eighteen months, and look instead at the underlying staging migration data — and a different picture emerges. One that three consecutive El Niño-influenced seasons have been quietly writing.
What the staging data is actually saying
IFRS 9's three-stage model was designed precisely for situations like this. Stage 1 loans carry a 12-month ECL. Stage 2 loans — where credit risk has significantly increased since origination — carry a lifetime ECL. Stage 3 is outright credit-impaired. The migration from Stage 1 to Stage 2 is the early warning signal most analysts underweight.
Across the SA banking sector, agri portfolio Stage 2 balances have grown materially over the past two reporting cycles. This is not yet reflected in Stage 3 NPL ratios — because many of these borrowers are still servicing interest, still rolling short-term facilities, still technically current. But the underlying cashflow capacity to service those facilities has eroded.
"The loan is current. The farmer is not."
The rainfall-indexed blind spot
The core problem is that standard IFRS 9 PD models were not built for rainfall-correlated credit risk. A commercial property loan defaults when the tenant stops paying. An agri loan defaults when the harvest fails — which is a function of rainfall, input costs, commodity price, and the farmer's working capital buffer. Standard through-the-cycle PD models, with qualitative overlays applied at year-end, are too blunt an instrument.
What's needed — and what almost no SA bank has yet implemented at portfolio level — is a rainfall-indexed PD overlay that dynamically adjusts ECL provisions based on seasonal forecasts. ENSO data is freely available. SAFEX futures price in commodity risk. The models can be built. The appetite to build them, before a regulator requires it, has been limited.
What Land Bank tells us
Land Bank is not a commercial bank. But its portfolio is the canary. It holds the highest concentration of smallholder and emerging farmer exposure in SA — precisely the borrower segment most vulnerable to rainfall shocks and least able to absorb input cost inflation. Its operational difficulties over the past three years are not a one-off governance failure. They are a stress test result for what happens when an agri credit book encounters sustained climate and macro headwinds without adequate capital buffers or funding diversity.
The commercial banks watching Land Bank's distress as a distant problem should be watching it as a leading indicator instead.
The credit cycle implication
SA's agri credit cycle is likely twelve to eighteen months behind the broader retail credit cycle in terms of stress recognition. Retail NPLs peaked in late 2024. Agri NPLs are still in the stage-2 accumulation phase. The recognition event — when staging migrations convert to full NPL classification — is coming. The question for credit risk teams is whether their IFRS 9 models will catch it early enough to allow adequate provisioning, or whether they will be restating overlays after the fact.
The quiet default is already happening. The numbers just haven't caught up yet.