I analyze whether variation in rainfall risk played a role in the demographic transition. The hypothesis is that children constituted a buffer stock of labor that could be mobilized in response to income shocks. Identification relies on fertility differences between farm and non-farm households within counties and over time. The results suggest that in areas with a high variance in rainfall the fertility differential was significantly higher than in areas with a low variance in rainfall. This channel is robust to other relevant forces and the spatial correlation in fertility. The effect disappeared as irrigation systems and agricultural machinery emerged.