Abstract The rainfall pattern in the Sahel is very erratic with a high spatial variability. We tested the often reported hypothesis that the dispersion of farmers’ fields around the village territory helps mitigate agro-climatic risk by increasing yield stability from year to year. We also wished to evaluate whether this strategy had an effect on the yield disparity among households in a village. Based on a network of approximately 60 rain gauges spread over 500 km 2 in the Fakara region (Southwest Niger), daily rainfall was interpolated at 300 m × 300 m resolution over a 12-year period. This data was used to compute, by means of the APSIM crop simulation model, millet biomass and grain yields at the pixel scale. Simulated yields were combined with the land tenure map of the Banizoumbou village in a GIS to assess millet yield at field and household level. Agro-climatic risk analysis was performed using linear regression between a spatial dispersion index of household fields and the inter-annual (instability) and inter-household (disparity) millet yield variability of 107 households in the village territory. We find that the spatial variability of annual rainfall induces an even higher spatial variability of millet production at pixel, field and household levels. The dispersion of farm fields reduces moderately but significantly the disparity of millet yield between households each year and increases the inter-annual yield stability of a given household. The less the household fields are scattered, the more the presence of a fertility gradient around the village enhances the inter-annual stability but also the disparity between households. Our results provide evidence that field dispersion is an effective strategy to mitigate agro-climatic risk, as claimed by farmers in the Sahelian Niger. Although the results should be confirmed by further research on longer term rainfall spatial data, it is clearly advisable that any land reforms in the area take into account the benefits of field dispersion to mitigate climatic risk.