An insurance contract for Malian cotton farmers used a number the buyer already had on file, instead of paying to collect a new one

the problem

Regional-average index insurance routinely fails to pay out even when a farmer's own village genuinely suffered a bad harvest

background

what everyone would do

Choose between a coarse regional index, cheap to run but prone to missing real localized losses, or a fine-grained village-level index that catches those losses but is cheap to game through local collusion -- the standard tradeoff every index-insurance designer faces, forcing a choice between basis risk and moral hazard rather than escaping it.

what they saw

The finer-grained number the village leaders wanted didn't need to be collected from scratch -- the cotton buyer was already recording it as an ordinary byproduct of grading and paying for the crop. Once that existing village-level figure was available for free, it could be used as the primary trigger while the zone-level index, instead of standing alone, could do a narrower job it was actually suited to: confirming a low village reading was genuine hardship and not a handful of neighbors colluding.

the move

Instead of choosing between the coarse-but-safe zone index and the fine-but-collusion-prone village index, the research team combined them using data that already existed for an unrelated purpose: Compagnie Malienne des Textiles (CMDT), the parastatal that buys all of Mali's cotton, already recorded village cooperative-level yields as part of its routine crop-grading and procurement bookkeeping. The resulting 'multiscale' contract set its primary payout trigger at the village level (yields below a threshold set between 264 and 913 kg/ha depending on the village, keeping premiums constant), but made that payout conditional on a secondary 'audit' check: the broader ZPA-level index also had to fall below its own threshold (900 kg/ha), confirming a low village yield reflected genuine misfortune rather than local collusion, without commissioning any new data collection to run that check.

why it works

Using the village-level yield the parastatal already recorded as the primary trigger makes the contract sensitive to genuinely local losses that a zone-wide average would wash out, directly attacking the false-negative problem farmers had identified. Requiring the broader zone index to also cross its own threshold before paying out gives insurers the moral-hazard protection a village-only contract couldn't offer, because a handful of colluding households in one cooperative can't move the yield average across an entire ten-village zone the same way they could manipulate their own cooperative's reported number. Because the zone-level data required for that check already existed too, the entire fix cost nothing to implement beyond combining two numbers the parastatal was recording anyway, which is why the false-negative rate fell from 45% to 7% at an unchanged premium rather than at a higher price reflecting new data-collection costs.

the payoff

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. Launched in 2011 across 86 cooperatives in the Bougouni region (two-thirds treatment, one-third control), 16 of the 58 treatment cooperatives (30%) purchased the multiscale contract in its first year -- well above the roughly 5% uptake a comparable single-trigger area-yield contract had drawn in a similar pilot in Peru, and consistent with simulations predicting about 40% higher demand than an equivalently priced single-scale contract.

where it breaks

The technique depends on a genuinely independent, hard-to-manipulate finer-grained data source already existing somewhere in the system -- if no such record exists, or if the same actor who might collude at the fine-grained level also controls the record itself, the combined signal offers no real protection against fraud. It also requires the coarse index to be a meaningfully different, harder-to-game measurement than the fine one, so that requiring both to agree actually screens out collusion rather than simply making payouts rarer across the board; the pilot itself was cut short by Mali's 2012 political instability before its long-run durability at scale could be fully tested.

what came after

Published as "Managing Basis Risk with Multi-scale Index Insurance Contracts" (Agricultural Economics 44(4-5), 2013), the pilot was cut short when Mali's March 2012 military coup destabilized many of the country's institutions; the research team moved the design to neighboring Burkina Faso, where a second version of the multiscale contract launched in 2013 under similar cotton-market conditions.

references

1. [1]Managing Basis Risk with Multiscale Index InsuranceMarc F. Bellemare (co-author, University of Minnesota), 2013marcfbellemare.com
2. [2]Managing basis risk with multiscale index insuranceUniversity of Minnesota Experts (institutional repository), 2013experts.umn.edu

keep it

same kind of clever

• 1991 · MarriottMarriott took the airlines' playbook and priced every room by the night it would die
• 2016 · NHS clinicsNHS clinics clawed back millions of missed appointments with reminder texts
• 1609 · Amsterdam Wisselbank (Bank of Amsterdam)Amsterdam stopped trying to police 800 coin types and just declared what a guilder meant instead