This study combines household survey data from Mali and Chad with ten rainfall datasets from various sources to examine how the choice of rainfall data influences estimates of drought impacts on food security. Rainfall shocks are measured using the Standardized Precipitation Index and assigned through bilinear, nearest-neighbor, and zonal aggregation techniques. Results without controls reveal substantial heterogeneity. Estimated coefficients vary widely in both magnitude and sign across datasets and interpolation methods, with coarse or satellite-only products exhibiting particularly high volatility. Once controls are added, this heterogeneity largely disappears and coefficients converge toward consistently negative values in line with theoretical expectations. Beyond magnitudes, the analysis also finds that the relative ranking of datasets shifts depending on the interpolation method and model specification, revealing instability in dataset performance. The findings highlight a fundamental challenge for applied economics since empirical conclusions may depend as much on dataset choice, spatial interpolation, and specification decisions as on the underlying relationship. This highlights the importance of transparency in the selection and processing of rainfall data.