Zambia insures a million farmers against drought. The cover was never built to protect a loan.

Since 2017, every farmer on Zambia's national input subsidy has been covered by crop index insurance, around a million people a season. After the 2023/24 drought the scheme paid out on a scale few programmes anywhere have matched, straight to farmers' mobile wallets. That is a real achievement, and it is worth being precise about what it does. The cover is sized to the subsidy and pays the farmer. It names no lender and is not sized to anything she has borrowed. So for a lender weighing whether to finance a smallholder's season, the district-wide drought remains the risk it cannot diversify away. Field experiments elsewhere suggest that when cover is tied to the loan, lenders approve more loans, even when farmers apply no more often. Zambia has already built the expensive half of that system: enrolment, satellite triggers and mobile payouts at national scale. The missing half is a record that connects a farmer's cover to what she owes.

The season that tested it

The 2023/24 season brought what Zambia's agricultural insurers describe as the worst drought in living memory.¹ It was also the season the country's subsidy-linked insurance was built for.

The World Food Programme, which has supported the scheme's design, reported that 722,000 of the 1,024,000 insured farmers received payouts worth just over ZMW 937 million, about US$34.4 million, and described it as the largest index payout to smallholder farmers from a single intervention.² Other parties report different totals. UNDP's 2026 case study puts the payout at over ZMW 800 million, around US$42.4 million, reaching 500,000 farmers; Hollard Zambia, one of the insurers, cites claims of over US$45 million.³¹ The sources do not reconcile, and we report the range rather than choose between them. On any of these figures the scale is unusual.

The money also moved differently from the past. Payouts went directly to farmers' mobile wallets through ZIAMIS, the Ministry of Agriculture's digital platform, which UNDP credits with removing delays that once came from channelling funds manually.³

The next season the rains recovered, and claims fell to under US$43,000.¹ That swing is not a flaw. It is what catastrophe cover is supposed to look like: very little in an ordinary year, a great deal in a bad one.

How the cover is built

In 2017 Cabinet made crop index insurance mandatory for every beneficiary of the Farmer Input Support Programme, funded from the farmer's deposit together with a government share.³ Insurers carry the risk and report when an index triggers; the Ministry then pays eligible farmers through ZIAMIS. Columbia University's International Research Institute for Climate and Society counts nearly four million cumulative farmers covered since the 2017/18 season.⁴

One design detail matters more than the rest. UNDP notes that at camp level the maximum payout is set roughly equal to the value of the input subsidy.³ The scheme protects the public investment in that farmer's inputs. It was asked to do that, and it does.

Three things follow, none of them a criticism. The sum insured tracks the subsidy, not the farmer's whole season or any debt she has taken on. The beneficiary is the farmer, paid to her own wallet, with no lender on the contract. And the trigger is an index measured over an area, not her field.

Why a lender sees no change

A lender deciding whether to finance a smallholder looks first at the risk it cannot manage. A single borrower's bad year can be absorbed with group liability, local knowledge and a sensible loan size. A drought that hits every borrower in a district in the same month cannot. That is systemic risk, and it is exactly the kind of risk index insurance is good at covering.

Under the current design, though, the drought payout arrives in the farmer's wallet, sized to her subsidy, with no link to what she owes. Whether any of it reaches a lender depends on choices a household makes in a hungry season, which is a reasonable thing for a household to decide and not something a credit officer can underwrite. The insurance exists. The lender cannot count on it.

What the experiments show

The most direct evidence comes from northern Ghana. In a randomised trial published in the American Journal of Agricultural Economics, farmer groups were offered production loans bundled with index insurance in one of two forms: cover that paid the farmer directly after a drought, or cover that paid the lender on condition the money retired the farmer's debt. A control group was offered ordinary loans. Insured loans raised farmers' likelihood of receiving credit by between 15 and 21 percentage points. The mechanism is the telling part: insured loans had no effect on whether farmers applied, but raised the likelihood of approval by between 17 and 25 points.⁵

The farmers did not change. The lenders did.

A second Ghana study, by Dean Karlan, Isaac Osei-Akoto, Robert Osei and Christopher Udry, points the same way from the farm side. Farmers given access to index insurance invested significantly more in their farms and took on riskier, higher-return production choices; the authors conclude that uninsured catastrophic risk, not only a shortage of cash, was holding investment back.⁶

The evidence is not one-sided, and the counter-case deserves its full weight. In Malawi, Xavier Giné and Dean Yang offered farmers loans for improved seed, some with compulsory weather insurance attached and paid for by the farmer. Take-up fell from 33 percent for the plain loan to 17.6 percent for the insured one. The authors' leading explanation is that the loan's limited liability already worked as informal insurance, so adding a premium simply made credit dearer. They also report that one lender later offered only insured loans without most borrowers knowing; the lender was safer, but the farmers' behaviour could not change because they did not know they were covered.⁷

The lesson is about design, not about whether insurance works. Who pays the premium, who is named as the beneficiary, and whether the farmer understands the cover all change the result.

The limits of the cover itself

Index insurance also has a well-known weakness: the index can miss a farmer's real loss. IRI's analysis for the Zambian programme found that the satellite weather index alone would cover about 30 to 40 percent of farmers' valued drought risk, and that district by district, fewer than half of farmers' worst years would have triggered a significant payout. Adding targeted yield checks where the weather index fails would raise coverage materially.⁴ The programme has since moved towards a hybrid of weather and area-yield cover, which its insurers say is meant to close exactly these gaps.¹³

For a lender this means index cover reduces systemic risk. It does not remove it, and any design that links cover to credit has to be honest about the share of bad years it will still miss.

What would link the two

Here we move from the evidence to our own reading, and we hold it as an interpretation rather than a proven result.

Linking drought cover to credit needs three things. Zambia already has two of them at national scale: a channel that enrols farmers and collects premiums, and a rail that pays them directly by mobile money when an index triggers. The product is improving. What is missing is a verifiable record that ties a specific farmer to a specific loan: what she borrowed, when, and what she has repaid. Without that record no insurer can size cover to her debt, and no programme can name a lender as loss payee for her, because the lender cannot show its book farmer by farmer. The Ghana design that paid the lender only worked because the lender knew exactly which debts the payout should retire.

If that reading is right, it has practical consequences for three groups.

For lenders and development finance institutions, drought cover can be built into agricultural credit lines as a condition of scale rather than an afterthought, and the first question is whether existing subsidy-linked cover can be topped up rather than duplicated.

For insurers and programme designers, the option worth testing is an optional layer above the subsidy-sized cover, available to farmers with a verified loan and naming the lender as loss payee for that layer only. The base cover, paid to the farmer, stays as it is. Malawi suggests farmers resist cover that feels like a surcharge on credit, so who pays for the top-up layer should be tested, not assumed.

For funders, the catalytic piece may be the credit record rather than the premium. Premium subsidies already exist. The connective data does not.

The strongest objection is that tying cover to debt could weaken a farmer's own protection at the moment she most needs cash, and that the scheme's reach came precisely from its simplicity. That is a serious concern, which is why we would keep any lender-linked layer optional, separate from the base cover, and small until it has been evaluated.

Where DNC sits

DNC lends farmers cash during the season, repaid in maize at harvest, and records every loan and purchase against the individual farmer through its platform, Panjiia. We do not offer insurance and we are not an insurer.

What our field experience has taught us is narrow. The hard part of connecting a farmer's risk to her credit is not the payout; Zambia has shown it can pay hundreds of thousands of farmers at once. It is knowing, verifiably, what each farmer owes and what she has repaid. That record is what we build.

Zambia has shown it can insure a million farmers and pay them after a drought. The next question is whether the same rails can help a lender lend to them before one.