Build1 publisher3 min readPublished
The $5m drought policy that paid nothing: Malawi's index was tuned to a crop nobody planted
Malawi's 2016 drought policy failed not because the automation broke but because the crop model was fitted to a maize variety farmers had abandoned. The check that would have caught it ran after the season.
The Engineer · Build desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction
What happened
- In April 2016 the government of Malawi declared a national emergency after rains failed across southern Africa for a second year running, with 6.7 million Malawians food-insecure and unable to feed themselves until the next harvest.
- For the 2015/16 season the government of Malawi paid almost US$5 million for a sovereign drought policy from ARC Ltd, the insurance affiliate of the African Risk Capacity, the African Union's disaster risk pool.
- The policy was parametric: instead of assessing losses on the ground, satellite rainfall estimates feed a crop model, the model estimates drought response costs, and when that estimate crosses a threshold money moves automatically, with no loss adjusters and no months of claims paperwork.
- The drought came and the emergency was declared, and the model concluded that no payout was warranted.
- The explanation most widely reported in the press, chiefly in Bloomberg's 2024 feature on Malawi and in coverage of ActionAid's 2017 study, is a single calibration choice: Africa RiskView, ARC's modelling platform, had been customised to track long-cycle maize, while most Malawian farmers had switched to planting short-cycle varieties.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
For the 2015/16 season the government of Malawi paid almost US$5 million to ARC Ltd, the insurance affiliate of the African Union's African Risk Capacity disaster risk pool, for a sovereign drought policy [2]. In April 2016 Malawi declared a national emergency with 6.7 million people unable to feed themselves until the next harvest [1], and the model concluded that no payout was warranted [4]. Premium in, US$5 million; payout out, zero [1].
This is worth studying because nothing in the pipeline malfunctioned. Parametric cover replaces loss adjustment with an index: satellite rainfall estimates feed a crop model, the model estimates drought response costs, and when the estimate crosses a threshold money moves automatically, with no loss adjusters and no months of claims paperwork [3]. In ARC's platform, Africa RiskView, rainfall feeds the Water Requirements Satisfaction Index, described as "an operational crop model originally developed by the United Nations Food and Agriculture Organization," using "information about crops, such as soil and cropping calendars," with each country selecting its own risk-transfer parameters [11]. The failure was upstream of all of that, in the fitting.
The explanation carried in the press, chiefly Bloomberg's 2024 feature on Malawi and coverage of ActionAid's 2017 study, is a single calibration choice: Africa RiskView had been customised to track long-cycle maize, while most Malawian farmers had switched to short-cycle varieties [5]. Those accounts also report that short-cycle hybrids are more sensitive to drought during flowering, not less [6], so the model was watching the hardier crop while the fragile one stood in the field. That account rests on press reporting. The 75-page independent evaluation commissioned after the crisis, by the e-Pact consortium led by Oxford Policy Management in October 2017 [8], never uses those terms [7].
What the evaluation does say is more useful to anyone running an automated payout system. It starts with basis risk, "the risk that there will be a mismatch between the payout that is triggered and the actual situation on the ground" [9]. That risk is the price of the speed, and it was not priced by accident: ARC describes customisation as a year-long process run through national technical working groups drawing on local expert knowledge [10]. Expert-staffed and year-long, and the evaluation still found it "does seem to have neglected adequate input from agronomists, agro-meteorologists and other critical expert stakeholders, and appears to have been too removed from the 'ground'" [12].
The verification step existed. Ground-truthing exercises in three districts in April and May 2016 "revealed discrepancies between ARV and realities on the ground" [13] - the same month the emergency was declared, by which point the season's outcome was already fixed [2]. Within a week ARC commissioned a consultant from Malawi's Centre for Agricultural Research and Development to investigate why the model had failed [14]. That post-mortem was never made public; the findings went to a stakeholder workshop in Malawi and the study itself was not released [15]. The one document that could confirm or complicate the maize story at source is unavailable.
The governance gap is the part that generalises. According to the evaluation, ARC "does not have a 'plain English' basis risk policy, and does not go through a formalised process to document and record agreement with countries that they understand the concept" [16]. A buyer signed for an instrument without a recorded understanding of how it could pay nothing in a disaster.
Three things to watch: whether the CARD post-mortem is ever published [15]; whether ground-truthing moves ahead of the trigger window rather than after it [13]; and whether contracts start carrying a documented, plain-language basis-risk acknowledgement from the buyer [16]. Until then, the failure mode is not code. It is a cropping calendar that stopped matching the fields.