Malian cotton farmers cut insurance false negatives using buyer's data

Index insurance pays out automatically when a regional yield or weather index crosses a bad-harvest threshold, avoiding the cost and fraud risk of verifying every individual farmer's loss. But a coarse regional index can stay near normal even while a specific village's own yields collapse, so the farmer who actually suffered a loss gets no payout -- a false-negative failure known as basis risk, and the single biggest reason index insurance struggles to attract buyers despite years of development-agency enthusiasm for it.

The standard fix -- shrinking the index to the village level, which leaders themselves proposed -- reintroduces the problem index insurance exists to avoid: a village cooperative of 20-30 closely connected households could plausibly collude to underreport yields or deliberately mismanage the crop to trigger a payout, a moral-hazard risk no insurer would accept at that scale.

Used the buyer's existing village yield records as the trigger, with the zone index as a collusion check, cutting false negatives from 45% to 7% at no extra cost.

Village-level trigger catches local losses a zone average washes out.

Zone-level audit check blocks collusion because a few households can't move a ten-village average.

Both data points already existed in the buyer's records, so the fix cost nothing extra.

Lower false negatives built trust, boosting uptake to 30% vs 5% in a comparable pilot.

Before paying for new data to reduce error, check if a finer-grained signal already exists in your records; combine it with the coarse one to catch cases the coarse one misses.

The two-index design cut the false-negative probability -- the chance a village suffering real losses received no payout -- from 45% under the conventional single-scale ZPA contract to just 7% at the same premium rate.