Index insurance, which indemnifies agricultural producers based on an objectively observable variable that is highly correlated with production losses but which cannot be influenced by the producer, can provide adequate protection against catastrophic droughts without suffering from the moral hazard and adverse selection problems that typically cause conventional agricultural insurance programs to fail. Using historical maize and cotton yield data from nine districts in Zimbabwe, we find that catastrophic drought insurance contract based on the Normalized Difference Vegetation Index (NDVI) can be constructed whose indemnities exhibit higher correlations with yield losses compared to the conventional rainfall index. In addition the NDVI contracts can be offered within the 5–10 per cent premium range considered reasonably affordable to many poor smallholder farmers in Zimbabwe.